generalize audit_del_rule

PATCH New Richard Guy Briggs 2 participants 4 messages Aug 1, 2015 – Aug 5, 2015 Audit

lore

Share & tools

lore

Richard Guy Briggs

author

2015-08-01 19:44 UTC #0 lore raw

PATCH v4 1/1 generalize audit_del_rule

This patch was split out from the audit by executable path patch set due to the potential to use it elsewhere.

In particular, some questions came up while assessing the potential for code reuse:

Why does audit_remove_parent_watches() not call audit_del_rule() for each entry found?
Is audit_signals not properly decremented?
Is audit_n_rules not properly decremented?

Why does kill_rules() not call audit_del_rule() for each entry found?
Is audit_signals not properly decremented?
Is audit_n_rules not properly decremented?

Richard Guy Briggs (1):
audit: save signal match info in case entry passed in is the one
deleted

 kernel/auditfilter.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

Richard Guy Briggs

author

2015-08-01 19:44 UTC #1 lore raw

PATCH v4 1/1 audit: save signal match info in case entry passed in is the one deleted

Move the access to the entry for audit_match_signal() to the beginning of the function in case the entry found is the same one passed in. This will enable it to be used by audit_remove_mark_rule().

Signed-off-by Richard Guy Briggs <[email protected]>

---
 kernel/auditfilter.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

kernel/auditfilter.c +2 -1

index 4cb9b44..afb63b3 100644
@@ -943,6 +943,7 @@ static inline int audit_del_rule(struct audit_entry *entry)
 	int ret = 0;
 #ifdef CONFIG_AUDITSYSCALL
 	int dont_count = 0;
+	int match_signal = !audit_match_signal(entry);
 
 	/* If either of these, don't count towards total */
 	if (entry->rule.listnr == AUDIT_FILTER_USER ||

@@ -972,7 +973,7 @@ static inline int audit_del_rule(struct audit_entry *entry)
 	if (!dont_count)
 		audit_n_rules--;
 
-	if (!audit_match_signal(entry))
+	if (match_signal)
 		audit_signals--;
 #endif
 	mutex_unlock(&audit_filter_mutex);

Signature · 1 line

1.7.1

Paul Moore

2015-08-04 23:04 UTC #2 lore raw

Re: [PATCH V4 (was V6)] audit: save signal match info in case entry passed in is the one deleted

Quoted 22 lines · Richard Guy Briggs

Move the access to the entry for audit_match_signal() to the beginning of the function in case the entry found is the same one passed in. This will enable it to be used by audit_remove_mark_rule().
Signed-off-by Richard Guy Briggs <[email protected]>
---
 kernel/auditfilter.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
kernel/auditfilter.c +1 -0
index 4cb9b44..afb63b3 100644
@@ -943,6 +943,7 @@ static inline int audit_del_rule(struct audit_entry
 *entry) int ret = 0;
 #ifdef CONFIG_AUDITSYSCALL
 	int dont_count = 0;
+	int match_signal = !audit_match_signal(entry);
 
 	/* If either of these, don't count towards total */
 	if (entry->rule.listnr == AUDIT_FILTER_USER ||

kernel/auditfilter.c static inline int audit_del_rule(struct audit_entry · line 979

Quoted 8 lines

kernel/auditfilter.c +1 -1

@@ -972,7 +973,7 @@ static inline int audit_del_rule(struct audit_entry
 *entry) if (!dont_count)
 		audit_n_rules--;
 
-	if (!audit_match_signal(entry))
+	if (match_signal)
 		audit_signals--;
 #endif
 	mutex_unlock(&audit_filter_mutex);

Why not simply move this second CONFIG_AUDITSYSCALL above the list_del() calls? Am I missing something?

Also, while we're fixing up audit_del_rule(), why not also move the mutex_unlock() call to after the "out" jump target and then drop the mutex_unlock() call in the audit_find_rule() error case? Not your fault, but the code seems silly as-is.

Signature · 2 lines

paul moore
security @ redhat

Richard Guy Briggs

author

2015-08-05 09:25 UTC #3 lore raw

Re: [PATCH V4 (was V6)] audit: save signal match info in case entry passed in is the one deleted

Quoted 26 lines · Paul Moore

Quoted 25 lines · Richard Guy Briggs
Move the access to the entry for audit_match_signal() to the beginning of the function in case the entry found is the same one passed in. This will enable it to be used by audit_remove_mark_rule().
Signed-off-by Richard Guy Briggs <[email protected]>
---
 kernel/auditfilter.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
kernel/auditfilter.c +1 -0
index 4cb9b44..afb63b3 100644
@@ -943,6 +943,7 @@ static inline int audit_del_rule(struct audit_entry
 *entry) int ret = 0;
 #ifdef CONFIG_AUDITSYSCALL
 	int dont_count = 0;
+	int match_signal = !audit_match_signal(entry);
 
 	/* If either of these, don't count towards total */
 	if (entry->rule.listnr == AUDIT_FILTER_USER ||
@@ -972,7 +973,7 @@ static inline int audit_del_rule(struct audit_entry
 *entry) if (!dont_count)
 		audit_n_rules--;
 

Quoted 8 lines

kernel/auditfilter.c static inline int audit_del_rule(struct audit_entry *entry) · line 979
Quoted 5 lines
kernel/auditfilter.c +1 -1
-	if (!audit_match_signal(entry))
+	if (match_signal)
 		audit_signals--;
 #endif
 	mutex_unlock(&audit_filter_mutex);
Why not simply move this second CONFIG_AUDITSYSCALL above the list_del() calls? Am I missing something?

Good point. That did occur to me at one point when I wasn't in front of the code and promptly forgot once I was. That will neatly remove the temporary variable.

Quoted 4 lines

Also, while we're fixing up audit_del_rule(), why not also move the mutex_unlock() call to after the "out" jump target and then drop the mutex_unlock() call in the audit_find_rule() error case? Not your fault, but the code seems silly as-is.

Yes, agreed. Another nice catch.

These changes will affect "audit by executable" so I'll re-spin that patch set to make it easier to apply.

Quoted 1 line

paul moore

- RGB

--
Richard Guy Briggs <[email protected]>
Senior Software Engineer, Kernel Security, AMER ENG Base Operating Systems, Red Hat Remote, Ottawa, Canada
Voice: +1.647.777.2635, Internal: (81) 32635, Alt: +1.613.693.0684x3545