2022-01-20 16:05:27 -05:00
|
|
|
start_server {tags {"acl external:skip"}} {
|
|
|
|
set r2 [redis_client]
|
|
|
|
test {Test basic multiple selectors} {
|
|
|
|
r ACL SETUSER selector-1 on -@all resetkeys nopass
|
|
|
|
$r2 auth selector-1 password
|
|
|
|
catch {$r2 ping} err
|
|
|
|
assert_match "*NOPERM*command*" $err
|
|
|
|
catch {$r2 set write::foo bar} err
|
|
|
|
assert_match "*NOPERM*command*" $err
|
|
|
|
catch {$r2 get read::foo} err
|
|
|
|
assert_match "*NOPERM*command*" $err
|
|
|
|
|
|
|
|
r ACL SETUSER selector-1 (+@write ~write::*) (+@read ~read::*)
|
|
|
|
catch {$r2 ping} err
|
|
|
|
assert_equal "OK" [$r2 set write::foo bar]
|
|
|
|
assert_equal "" [$r2 get read::foo]
|
|
|
|
catch {$r2 get write::foo} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 set read::foo bar} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
2022-01-30 05:02:55 -05:00
|
|
|
test {Test ACL selectors by default have no permissions} {
|
2022-01-20 16:05:27 -05:00
|
|
|
r ACL SETUSER selector-default reset ()
|
|
|
|
set user [r ACL GETUSER "selector-default"]
|
|
|
|
assert_equal 1 [llength [dict get $user selectors]]
|
|
|
|
assert_equal "" [dict get [lindex [dict get $user selectors] 0] keys]
|
2022-01-30 05:02:55 -05:00
|
|
|
assert_equal "" [dict get [lindex [dict get $user selectors] 0] channels]
|
2022-01-20 16:05:27 -05:00
|
|
|
assert_equal "-@all" [dict get [lindex [dict get $user selectors] 0] commands]
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test deleting selectors} {
|
|
|
|
r ACL SETUSER selector-del on "(~added-selector)"
|
|
|
|
set user [r ACL GETUSER "selector-del"]
|
|
|
|
assert_equal "~added-selector" [dict get [lindex [dict get $user selectors] 0] keys]
|
|
|
|
assert_equal [llength [dict get $user selectors]] 1
|
|
|
|
|
|
|
|
r ACL SETUSER selector-del clearselectors
|
|
|
|
set user [r ACL GETUSER "selector-del"]
|
|
|
|
assert_equal [llength [dict get $user selectors]] 0
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test selector syntax error reports the error in the selector context} {
|
|
|
|
catch {r ACL SETUSER selector-syntax on (this-is-invalid)} e
|
|
|
|
assert_match "*ERR Error in ACL SETUSER modifier '(*)*Syntax*" $e
|
|
|
|
|
2022-01-30 05:02:55 -05:00
|
|
|
catch {r ACL SETUSER selector-syntax on (&* &fail)} e
|
2022-01-20 16:05:27 -05:00
|
|
|
assert_match "*ERR Error in ACL SETUSER modifier '(*)*Adding a pattern after the*" $e
|
|
|
|
|
|
|
|
assert_equal "" [r ACL GETUSER selector-syntax]
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test flexible selector definition} {
|
|
|
|
# Test valid selectors
|
|
|
|
r ACL SETUSER selector-2 "(~key1 +get )" "( ~key2 +get )" "( ~key3 +get)" "(~key4 +get)"
|
|
|
|
r ACL SETUSER selector-2 (~key5 +get ) ( ~key6 +get ) ( ~key7 +get) (~key8 +get)
|
|
|
|
set user [r ACL GETUSER "selector-2"]
|
|
|
|
assert_equal "~key1" [dict get [lindex [dict get $user selectors] 0] keys]
|
|
|
|
assert_equal "~key2" [dict get [lindex [dict get $user selectors] 1] keys]
|
|
|
|
assert_equal "~key3" [dict get [lindex [dict get $user selectors] 2] keys]
|
|
|
|
assert_equal "~key4" [dict get [lindex [dict get $user selectors] 3] keys]
|
|
|
|
assert_equal "~key5" [dict get [lindex [dict get $user selectors] 4] keys]
|
|
|
|
assert_equal "~key6" [dict get [lindex [dict get $user selectors] 5] keys]
|
|
|
|
assert_equal "~key7" [dict get [lindex [dict get $user selectors] 6] keys]
|
|
|
|
assert_equal "~key8" [dict get [lindex [dict get $user selectors] 7] keys]
|
|
|
|
|
|
|
|
# Test invalid selector syntax
|
|
|
|
catch {r ACL SETUSER invalid-selector " () "} err
|
|
|
|
assert_match "*ERR*Syntax error*" $err
|
|
|
|
catch {r ACL SETUSER invalid-selector (} err
|
|
|
|
assert_match "*Unmatched parenthesis*" $err
|
|
|
|
catch {r ACL SETUSER invalid-selector )} err
|
|
|
|
assert_match "*ERR*Syntax error" $err
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test separate read permission} {
|
|
|
|
r ACL SETUSER key-permission-R on nopass %R~read* +@all
|
|
|
|
$r2 auth key-permission-R password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
r set readstr bar
|
|
|
|
assert_equal bar [$r2 get readstr]
|
|
|
|
catch {$r2 set readstr bar} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 get notread} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test separate write permission} {
|
|
|
|
r ACL SETUSER key-permission-W on nopass %W~write* +@all
|
|
|
|
$r2 auth key-permission-W password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
# Note, SET is a RW command, so it's not used for testing
|
|
|
|
$r2 LPUSH writelist 10
|
|
|
|
catch {$r2 GET writestr} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 LPUSH notwrite 10} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test separate read and write permissions} {
|
|
|
|
r ACL SETUSER key-permission-RW on nopass %R~read* %W~write* +@all
|
|
|
|
$r2 auth key-permission-RW password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
r set read bar
|
|
|
|
$r2 copy read write
|
|
|
|
catch {$r2 copy write read} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test separate read and write permissions on different selectors are not additive} {
|
|
|
|
r ACL SETUSER key-permission-RW-selector on nopass "(%R~read* +@all)" "(%W~write* +@all)"
|
|
|
|
$r2 auth key-permission-RW-selector password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
|
|
|
|
# Verify write selector
|
|
|
|
$r2 LPUSH writelist 10
|
|
|
|
catch {$r2 GET writestr} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 LPUSH notwrite 10} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
|
|
|
|
# Verify read selector
|
|
|
|
r set readstr bar
|
|
|
|
assert_equal bar [$r2 get readstr]
|
|
|
|
catch {$r2 set readstr bar} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 get notread} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
|
|
|
|
# Verify they don't combine
|
|
|
|
catch {$r2 copy read write} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
catch {$r2 copy write read} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-26 14:03:21 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test SET with separate read permission} {
|
|
|
|
r del readstr
|
|
|
|
r ACL SETUSER set-key-permission-R on nopass %R~read* +@all
|
|
|
|
$r2 auth set-key-permission-R password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
assert_equal {} [$r2 get readstr]
|
|
|
|
|
|
|
|
# We don't have the permission to WRITE key.
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_error {*NOPERM*key*} {$r2 set readstr bar}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set readstr bar get}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set readstr bar ex 100}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set readstr bar keepttl nx}
|
2022-01-26 14:03:21 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test SET with separate write permission} {
|
|
|
|
r del writestr
|
|
|
|
r ACL SETUSER set-key-permission-W on nopass %W~write* +@all
|
|
|
|
$r2 auth set-key-permission-W password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
assert_equal {OK} [$r2 set writestr bar]
|
|
|
|
assert_equal {OK} [$r2 set writestr get]
|
|
|
|
|
|
|
|
# We don't have the permission to READ key.
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_error {*NOPERM*key*} {$r2 set get writestr}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set writestr bar get}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set writestr bar get ex 100}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 set writestr bar get keepttl nx}
|
2022-01-26 14:03:21 -05:00
|
|
|
|
|
|
|
# this probably should be `ERR value is not an integer or out of range`
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_error {*NOPERM*key*} {$r2 set writestr bar ex get}
|
2022-01-26 14:03:21 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test SET with read and write permissions} {
|
|
|
|
r del readwrite_str
|
|
|
|
r ACL SETUSER set-key-permission-RW-selector on nopass %RW~readwrite* +@all
|
|
|
|
$r2 auth set-key-permission-RW-selector password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
|
|
|
|
assert_equal {} [$r2 get readwrite_str]
|
2022-04-14 04:18:32 -04:00
|
|
|
assert_error {ERR * not an integer *} {$r2 set readwrite_str bar ex get}
|
2022-01-26 14:03:21 -05:00
|
|
|
|
|
|
|
assert_equal {OK} [$r2 set readwrite_str bar]
|
|
|
|
assert_equal {bar} [$r2 get readwrite_str]
|
|
|
|
|
|
|
|
assert_equal {bar} [$r2 set readwrite_str bar2 get]
|
|
|
|
assert_equal {bar2} [$r2 get readwrite_str]
|
|
|
|
|
|
|
|
assert_equal {bar2} [$r2 set readwrite_str bar3 get ex 10]
|
|
|
|
assert_equal {bar3} [$r2 get readwrite_str]
|
|
|
|
assert_range [$r2 ttl readwrite_str] 5 10
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test BITFIELD with separate read permission} {
|
|
|
|
r del readstr
|
|
|
|
r ACL SETUSER bitfield-key-permission-R on nopass %R~read* +@all
|
|
|
|
$r2 auth bitfield-key-permission-R password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
assert_equal {0} [$r2 bitfield readstr get u4 0]
|
|
|
|
|
|
|
|
# We don't have the permission to WRITE key.
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield readstr set u4 0 1}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield readstr get u4 0 set u4 0 1}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield readstr incrby u4 0 1}
|
2022-01-26 14:03:21 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test BITFIELD with separate write permission} {
|
|
|
|
r del writestr
|
|
|
|
r ACL SETUSER bitfield-key-permission-W on nopass %W~write* +@all
|
|
|
|
$r2 auth bitfield-key-permission-W password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
|
|
|
|
# We don't have the permission to READ key.
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield writestr get u4 0}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield writestr set u4 0 1}
|
|
|
|
assert_error {*NOPERM*key*} {$r2 bitfield writestr incrby u4 0 1}
|
2022-01-26 14:03:21 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test BITFIELD with read and write permissions} {
|
|
|
|
r del readwrite_str
|
|
|
|
r ACL SETUSER bitfield-key-permission-RW-selector on nopass %RW~readwrite* +@all
|
|
|
|
$r2 auth bitfield-key-permission-RW-selector password
|
|
|
|
assert_equal PONG [$r2 PING]
|
|
|
|
|
|
|
|
assert_equal {0} [$r2 bitfield readwrite_str get u4 0]
|
|
|
|
assert_equal {0} [$r2 bitfield readwrite_str set u4 0 1]
|
|
|
|
assert_equal {2} [$r2 bitfield readwrite_str incrby u4 0 1]
|
|
|
|
assert_equal {2} [$r2 bitfield readwrite_str get u4 0]
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test ACL log correctly identifies the relevant item when selectors are used} {
|
2022-01-26 14:03:21 -05:00
|
|
|
r ACL SETUSER acl-log-test-selector on nopass
|
2022-01-20 16:05:27 -05:00
|
|
|
r ACL SETUSER acl-log-test-selector +mget ~key (+mget ~key ~otherkey)
|
|
|
|
$r2 auth acl-log-test-selector password
|
|
|
|
|
|
|
|
# Test that command is shown only if none of the selectors match
|
|
|
|
r ACL LOG RESET
|
|
|
|
catch {$r2 GET key} err
|
|
|
|
assert_match "*NOPERM*command*" $err
|
|
|
|
set entry [lindex [r ACL LOG] 0]
|
|
|
|
assert_equal [dict get $entry username] "acl-log-test-selector"
|
|
|
|
assert_equal [dict get $entry context] "toplevel"
|
|
|
|
assert_equal [dict get $entry reason] "command"
|
|
|
|
assert_equal [dict get $entry object] "get"
|
|
|
|
|
|
|
|
# Test two cases where the first selector matches less than the
|
|
|
|
# second selector. We should still show the logically first unmatched key.
|
|
|
|
r ACL LOG RESET
|
|
|
|
catch {$r2 MGET otherkey someotherkey} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
set entry [lindex [r ACL LOG] 0]
|
|
|
|
assert_equal [dict get $entry username] "acl-log-test-selector"
|
|
|
|
assert_equal [dict get $entry context] "toplevel"
|
|
|
|
assert_equal [dict get $entry reason] "key"
|
|
|
|
assert_equal [dict get $entry object] "someotherkey"
|
|
|
|
|
|
|
|
r ACL LOG RESET
|
|
|
|
catch {$r2 MGET key otherkey someotherkey} err
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*NOPERM*key*" $err
|
2022-01-20 16:05:27 -05:00
|
|
|
set entry [lindex [r ACL LOG] 0]
|
|
|
|
assert_equal [dict get $entry username] "acl-log-test-selector"
|
|
|
|
assert_equal [dict get $entry context] "toplevel"
|
|
|
|
assert_equal [dict get $entry reason] "key"
|
|
|
|
assert_equal [dict get $entry object] "someotherkey"
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test ACL GETUSER response information} {
|
2022-01-26 14:03:21 -05:00
|
|
|
r ACL setuser selector-info -@all +get resetchannels &channel1 %R~foo1 %W~bar1 ~baz1
|
2022-01-20 16:05:27 -05:00
|
|
|
r ACL setuser selector-info (-@all +set resetchannels &channel2 %R~foo2 %W~bar2 ~baz2)
|
|
|
|
set user [r ACL GETUSER "selector-info"]
|
2022-01-26 14:03:21 -05:00
|
|
|
|
2022-01-20 16:05:27 -05:00
|
|
|
# Root selector
|
|
|
|
assert_equal "%R~foo1 %W~bar1 ~baz1" [dict get $user keys]
|
|
|
|
assert_equal "&channel1" [dict get $user channels]
|
|
|
|
assert_equal "-@all +get" [dict get $user commands]
|
|
|
|
|
|
|
|
# Added selector
|
|
|
|
set secondary_selector [lindex [dict get $user selectors] 0]
|
|
|
|
assert_equal "%R~foo2 %W~bar2 ~baz2" [dict get $secondary_selector keys]
|
|
|
|
assert_equal "&channel2" [dict get $secondary_selector channels]
|
2022-01-26 14:03:21 -05:00
|
|
|
assert_equal "-@all +set" [dict get $secondary_selector commands]
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test ACL list idempotency} {
|
|
|
|
r ACL SETUSER user-idempotency off -@all +get resetchannels &channel1 %R~foo1 %W~bar1 ~baz1 (-@all +set resetchannels &channel2 %R~foo2 %W~bar2 ~baz2)
|
|
|
|
set response [lindex [r ACL LIST] [lsearch [r ACL LIST] "user user-idempotency*"]]
|
|
|
|
|
|
|
|
assert_match "*-@all*+get*(*)*" $response
|
|
|
|
assert_match "*resetchannels*&channel1*(*)*" $response
|
|
|
|
assert_match "*%R~foo1*%W~bar1*~baz1*(*)*" $response
|
|
|
|
|
|
|
|
assert_match "*(*-@all*+set*)*" $response
|
|
|
|
assert_match "*(*resetchannels*&channel2*)*" $response
|
|
|
|
assert_match "*(*%R~foo2*%W~bar2*~baz2*)*" $response
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test R+W is the same as all permissions} {
|
|
|
|
r ACL setuser selector-rw-info %R~foo %W~foo %RW~bar
|
|
|
|
set user [r ACL GETUSER selector-rw-info]
|
|
|
|
assert_equal "~foo ~bar" [dict get $user keys]
|
|
|
|
}
|
|
|
|
|
|
|
|
test {Test basic dry run functionality} {
|
|
|
|
r ACL setuser command-test +@all %R~read* %W~write* %RW~rw*
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test GET read]
|
|
|
|
|
|
|
|
catch {r ACL DRYRUN not-a-user GET read} e
|
|
|
|
assert_equal "ERR User 'not-a-user' not found" $e
|
|
|
|
|
|
|
|
catch {r ACL DRYRUN command-test not-a-command read} e
|
|
|
|
assert_equal "ERR Command 'not-a-command' not found" $e
|
|
|
|
}
|
|
|
|
|
2022-02-28 23:26:58 -05:00
|
|
|
test {Test various commands for command permissions} {
|
|
|
|
r ACL setuser command-test -@all
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to run the 'set' command*} [r ACL DRYRUN command-test set somekey somevalue]
|
|
|
|
assert_match {*has no permissions to run the 'get' command*} [r ACL DRYRUN command-test get somekey]
|
2022-02-28 23:26:58 -05:00
|
|
|
}
|
|
|
|
|
2022-01-20 16:05:27 -05:00
|
|
|
test {Test various odd commands for key permissions} {
|
|
|
|
r ACL setuser command-test +@all %R~read* %W~write* %RW~rw*
|
|
|
|
|
|
|
|
# Test migrate, which is marked with incomplete keys
|
2022-03-10 03:08:41 -05:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MIGRATE whatever whatever rw 0 500]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'read' key*} [r ACL DRYRUN command-test MIGRATE whatever whatever read 0 500]
|
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test MIGRATE whatever whatever write 0 500]
|
2022-01-20 16:05:27 -05:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 KEYS rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*has no permissions to access the 'read' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 KEYS read]
|
|
|
|
assert_match "*has no permissions to access the 'write' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 KEYS write]
|
2022-10-13 08:03:54 -04:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH KEYS KEYS rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*has no permissions to access the 'read' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH KEYS KEYS read]
|
|
|
|
assert_match "*has no permissions to access the 'write' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH KEYS KEYS write]
|
2022-10-13 08:03:54 -04:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 KEYS 123 KEYS rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*has no permissions to access the 'read' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 KEYS 123 KEYS read]
|
|
|
|
assert_match "*has no permissions to access the 'write' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 KEYS 123 KEYS write]
|
2022-10-13 08:03:54 -04:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 USER KEYS KEYS rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match "*has no permissions to access the 'read' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 USER KEYS KEYS read]
|
|
|
|
assert_match "*has no permissions to access the 'write' key" [r ACL DRYRUN command-test MIGRATE whatever whatever "" 0 5000 AUTH2 USER KEYS KEYS write]
|
2022-01-20 16:05:27 -05:00
|
|
|
|
|
|
|
# Test SORT, which is marked with incomplete keys
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test SORT read STORE write]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'read' key*} [r ACL DRYRUN command-test SORT read STORE read]
|
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test SORT write STORE write]
|
2022-01-20 16:05:27 -05:00
|
|
|
|
|
|
|
# Test EVAL, which uses the numkey keyspec (Also test EVAL_RO)
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL "" 1 rw1]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'read' key*} [r ACL DRYRUN command-test EVAL "" 1 read]
|
2022-01-20 16:05:27 -05:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL_RO "" 1 rw1]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL_RO "" 1 read]
|
|
|
|
|
|
|
|
# Read is an optional argument and not a key here, make sure we don't treat it as a key
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL "" 0 read]
|
|
|
|
|
|
|
|
# These are syntax errors, but it's 'OK' from an ACL perspective
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL "" -1 read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL "" 3 rw rw]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test EVAL "" 3 rw read]
|
|
|
|
|
|
|
|
# Test GEORADIUS which uses the last type of keyspec, keyword
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test GEORADIUS read longitude latitude radius M STOREDIST write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test GEORADIUS read longitude latitude radius M]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'read2' key*} [r ACL DRYRUN command-test GEORADIUS read1 longitude latitude radius M STOREDIST read2]
|
|
|
|
assert_match {*has no permissions to access the 'write1' key*} [r ACL DRYRUN command-test GEORADIUS write1 longitude latitude radius M STOREDIST write2]
|
2022-01-20 16:05:27 -05:00
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test GEORADIUS read longitude latitude radius M STORE write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test GEORADIUS read longitude latitude radius M]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'read2' key*} [r ACL DRYRUN command-test GEORADIUS read1 longitude latitude radius M STORE read2]
|
|
|
|
assert_match {*has no permissions to access the 'write1' key*} [r ACL DRYRUN command-test GEORADIUS write1 longitude latitude radius M STORE write2]
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
|
|
|
|
2022-01-25 02:55:30 -05:00
|
|
|
# Existence test commands are not marked as access since they are the result
|
|
|
|
# of a lot of write commands. We therefore make the claim they can be executed
|
|
|
|
# when either READ or WRITE flags are provided.
|
|
|
|
test {Existence test commands are not marked as access} {
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test HEXISTS read foo]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test HEXISTS write foo]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test HEXISTS nothing foo]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test HSTRLEN read foo]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test HSTRLEN write foo]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test HSTRLEN nothing foo]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test SISMEMBER read foo]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test SISMEMBER write foo]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test SISMEMBER nothing foo]
|
2022-01-25 02:55:30 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
# Unlike existence test commands, intersection cardinality commands process the data
|
|
|
|
# between keys and return an aggregated cardinality. therefore they have the access
|
|
|
|
# requirement.
|
|
|
|
test {Intersection cardinaltiy commands are access commands} {
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test SINTERCARD 2 read read]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test SINTERCARD 2 write read]
|
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test SINTERCARD 2 nothing read]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test ZCOUNT read 0 1]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test ZCOUNT write 0 1]
|
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test ZCOUNT nothing 0 1]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test PFCOUNT read read]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test PFCOUNT write read]
|
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test PFCOUNT nothing read]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test ZINTERCARD 2 read read]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'write' key*} [r ACL DRYRUN command-test ZINTERCARD 2 write read]
|
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test ZINTERCARD 2 nothing read]
|
2022-01-25 02:55:30 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Test general keyspace commands require some type of permission to execute} {
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test touch read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test touch write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test touch rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test touch nothing]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test exists read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test exists write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test exists rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test exists nothing]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MEMORY USAGE read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MEMORY USAGE write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test MEMORY USAGE rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test MEMORY USAGE nothing]
|
2022-01-25 02:55:30 -05:00
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test TYPE read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test TYPE write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test TYPE rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test TYPE nothing]
|
2022-01-25 02:55:30 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
test {Cardinality commands require some type of permission to execute} {
|
|
|
|
set commands {STRLEN HLEN LLEN SCARD ZCARD XLEN}
|
|
|
|
foreach command $commands {
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test $command read]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test $command write]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN command-test $command rw]
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'nothing' key*} [r ACL DRYRUN command-test $command nothing]
|
2022-01-25 02:55:30 -05:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-01-20 16:05:27 -05:00
|
|
|
test {Test sharded channel permissions} {
|
|
|
|
r ACL setuser test-channels +@all resetchannels &channel
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-channels spublish channel foo]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-channels ssubscribe channel]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-channels sunsubscribe]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-channels sunsubscribe channel]
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-channels sunsubscribe otherchannel]
|
|
|
|
|
Unify ACL failure error messaging. (#11160)
Motivation: for applications that use RM ACL verification functions, they would
want to return errors back to the user, in ways that are consistent with Redis.
While investigating how we should return ACL errors to the user, we realized that
Redis isn't consistent, and currently returns ACL error strings in 3 primary ways.
[For the actual implications of this change, see the "Impact" section at the bottom]
1. how it returns an error when calling a command normally
ACL_DENIED_CMD -> "this user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "this user has no permissions to access one of the keys used as arguments"
ACL_DENIED_CHANNEL -> "this user has no permissions to access one of the channels used as arguments"
2. how it returns an error when calling via 'acl dryrun' command
ACL_DENIED_CMD -> "This user has no permissions to run the '%s' command"
ACL_DENIED_KEY -> "This user has no permissions to access the '%s' key"
ACL_DENIED_CHANNEL -> "This user has no permissions to access the '%s' channel"
3. how it returns an error via RM_Call (and scripting is similar).
ACL_DENIED_CMD -> "can't run this command or subcommand";
ACL_DENIED_KEY -> "can't access at least one of the keys mentioned in the command arguments";
ACL_DENIED_CHANNEL -> "can't publish to the channel mentioned in the command";
In addition, if one wants to use RM_Call's "dry run" capability instead of the RM ACL
functions directly, one also sees a different problem than it returns ACL errors with a -ERR,
not a -PERM, so it can't be returned directly to the caller.
This PR modifies the code to generate a base message in a common manner with the ability
to set verbose flag for acl dry run errors, and keep it unset for normal/rm_call/script cases
```c
sds getAclErrorMessage(int acl_res, user *user, struct redisCommand *cmd, sds errored_val, int verbose) {
switch (acl_res) {
case ACL_DENIED_CMD:
return sdscatfmt(sdsempty(), "User %S has no permissions to run "
"the '%S' command", user->name, cmd->fullname);
case ACL_DENIED_KEY:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' key", user->name, errored_val);
} else {
return sdsnew("No permissions to access a key");
}
case ACL_DENIED_CHANNEL:
if (verbose) {
return sdscatfmt(sdsempty(), "User %S has no permissions to access "
"the '%S' channel", user->name, errored_val);
} else {
return sdsnew("No permissions to access a channel");
}
}
```
The caller can append/prepend the message (adding NOPERM for normal/RM_Call or indicating it's within a script).
Impact:
- Plain commands, as well as scripts and RM_Call now include the user name.
- ACL DRYRUN remains the only one that's verbose (mentions the offending channel or key name)
- Changes RM_Call ACL errors from being a `-ERR` to being `-NOPERM` (besides for textual changes)
**This somewhat a breaking change, but it only affects the RM_Call with both `C` and `E`, or `D`**
- Changes ACL errors in scripts textually from being
`The user executing the script <old non unified text>`
to
`ACL failure in script: <new unified text>`
2022-10-16 02:01:37 -04:00
|
|
|
assert_match {*has no permissions to access the 'otherchannel' channel*} [r ACL DRYRUN test-channels spublish otherchannel foo]
|
|
|
|
assert_match {*has no permissions to access the 'otherchannel' channel*} [r ACL DRYRUN test-channels ssubscribe otherchannel foo]
|
2022-01-20 16:05:27 -05:00
|
|
|
}
|
2022-03-15 11:14:53 -04:00
|
|
|
|
|
|
|
test {Test sort with ACL permissions} {
|
|
|
|
r set v1 1
|
|
|
|
r lpush mylist 1
|
|
|
|
|
|
|
|
r ACL setuser test-sort-acl on nopass (+sort ~mylist)
|
|
|
|
$r2 auth test-sort-acl nopass
|
|
|
|
|
|
|
|
catch {$r2 sort mylist by v*} e
|
|
|
|
assert_equal "ERR BY option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
catch {$r2 sort mylist get v*} e
|
|
|
|
assert_equal "ERR GET option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
|
|
|
|
r ACL setuser test-sort-acl (+sort ~mylist ~v*)
|
|
|
|
catch {$r2 sort mylist by v*} e
|
|
|
|
assert_equal "ERR BY option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
catch {$r2 sort mylist get v*} e
|
|
|
|
assert_equal "ERR GET option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
|
|
|
|
r ACL setuser test-sort-acl (+sort ~mylist %W~*)
|
|
|
|
catch {$r2 sort mylist by v*} e
|
|
|
|
assert_equal "ERR BY option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
catch {$r2 sort mylist get v*} e
|
|
|
|
assert_equal "ERR GET option of SORT denied due to insufficient ACL permissions." $e
|
|
|
|
|
|
|
|
r ACL setuser test-sort-acl (+sort ~mylist %R~*)
|
|
|
|
assert_equal "1" [$r2 sort mylist by v*]
|
|
|
|
|
|
|
|
# cleanup
|
|
|
|
r ACL deluser test-sort-acl
|
|
|
|
r del v1 mylist
|
|
|
|
}
|
2022-03-10 03:08:41 -05:00
|
|
|
|
|
|
|
test {Test DRYRUN with wrong number of arguments} {
|
|
|
|
r ACL setuser test-dry-run +@all ~v*
|
|
|
|
|
|
|
|
assert_equal "OK" [r ACL DRYRUN test-dry-run SET v v]
|
|
|
|
|
|
|
|
catch {r ACL DRYRUN test-dry-run SET v} e
|
|
|
|
assert_equal "ERR wrong number of arguments for 'set' command" $e
|
|
|
|
|
|
|
|
catch {r ACL DRYRUN test-dry-run SET} e
|
|
|
|
assert_equal "ERR wrong number of arguments for 'set' command" $e
|
|
|
|
}
|
2022-01-20 16:05:27 -05:00
|
|
|
|
|
|
|
$r2 close
|
|
|
|
}
|
|
|
|
|
|
|
|
set server_path [tmpdir "selectors.acl"]
|
|
|
|
exec cp -f tests/assets/userwithselectors.acl $server_path
|
|
|
|
exec cp -f tests/assets/default.conf $server_path
|
|
|
|
start_server [list overrides [list "dir" $server_path "aclfile" "userwithselectors.acl"] tags [list "external:skip"]] {
|
|
|
|
|
|
|
|
test {Test behavior of loading ACLs} {
|
|
|
|
set selectors [dict get [r ACL getuser alice] selectors]
|
|
|
|
assert_equal [llength $selectors] 1
|
|
|
|
set test_selector [lindex $selectors 0]
|
|
|
|
assert_equal "-@all +get" [dict get $test_selector "commands"]
|
|
|
|
assert_equal "~rw*" [dict get $test_selector "keys"]
|
|
|
|
|
|
|
|
set selectors [dict get [r ACL getuser bob] selectors]
|
|
|
|
assert_equal [llength $selectors] 2
|
|
|
|
set test_selector [lindex $selectors 0]
|
|
|
|
assert_equal "-@all +set" [dict get $test_selector "commands"]
|
|
|
|
assert_equal "%W~w*" [dict get $test_selector "keys"]
|
|
|
|
|
|
|
|
set test_selector [lindex $selectors 1]
|
|
|
|
assert_equal "-@all +get" [dict get $test_selector "commands"]
|
|
|
|
assert_equal "%R~r*" [dict get $test_selector "keys"]
|
|
|
|
}
|
|
|
|
}
|