* [PATCH] audit: fix a memleak caused by auditing load module @ 2019-03-05 11:14 Li RongQing 2019-03-05 14:18 ` Paul Moore 0 siblings, 1 reply; 3+ messages in thread From: Li RongQing @ 2019-03-05 11:14 UTC (permalink / raw) To: paul, eparis, linux-audit, linux-kernel we should always free context->module.name, since it will be allocated unconditionally and audit_log_start() can fail with other reasons, and audit_log_exit maybe not called unreferenced object 0xffff88af90837d20 (size 8): comm "modprobe", pid 1036, jiffies 4294704867 (age 3069.138s) hex dump (first 8 bytes): 69 78 67 62 65 00 ff ff ixgbe... backtrace: [<0000000008da28fe>] __audit_log_kern_module+0x33/0x80 [<00000000c1491e61>] load_module+0x64f/0x3850 [<000000007fc9ae3f>] __do_sys_init_module+0x218/0x250 [<0000000000d4a478>] do_syscall_64+0x117/0x400 [<000000004924ded8>] entry_SYSCALL_64_after_hwframe+0x49/0xbe [<000000007dc331dd>] 0xffffffffffffffff Fixes: ca86cad7380e3 ("audit: log module name on init_module") Signed-off-by: Zhang Yu <zhangyu31@baidu.com> Signed-off-by: Li RongQing <lirongqing@baidu.com> --- kernel/auditsc.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) diff --git a/kernel/auditsc.c b/kernel/auditsc.c index b2d1f043f..2bd80375f 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -1186,8 +1186,13 @@ static void show_special(struct audit_context *context, int *call_panic) int i; ab = audit_log_start(context, GFP_KERNEL, context->type); - if (!ab) + if (!ab) { + if (context->type == AUDIT_KERN_MODULE) { + kfree(context->module.name); + context->module.name = NULL; + } return; + } switch (context->type) { case AUDIT_SOCKETCALL: { @@ -1354,8 +1359,15 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts context->personality = tsk->personality; ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL); - if (!ab) + + if (!ab) { + if (context->type == AUDIT_KERN_MODULE) { + kfree(context->module.name); + context->module.name = NULL; + } return; /* audit_panic has been called */ + } + audit_log_format(ab, "arch=%x syscall=%d", context->arch, context->major); if (context->personality != PER_LINUX) @@ -1576,6 +1588,12 @@ void __audit_syscall_exit(int success, long return_code) if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT) audit_log_exit(context, current); + else { + if (context->type == AUDIT_KERN_MODULE) { + kfree(context->module.name); + context->module.name = NULL; + } + } context->in_syscall = 0; context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0; -- 2.16.2 ^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] audit: fix a memleak caused by auditing load module 2019-03-05 11:14 [PATCH] audit: fix a memleak caused by auditing load module Li RongQing @ 2019-03-05 14:18 ` Paul Moore 2019-03-06 3:23 ` 答复: " Li,Rongqing 0 siblings, 1 reply; 3+ messages in thread From: Paul Moore @ 2019-03-05 14:18 UTC (permalink / raw) To: Li RongQing; +Cc: Eric Paris, linux-audit, linux-kernel On Tue, Mar 5, 2019 at 6:14 AM Li RongQing <lirongqing@baidu.com> wrote: > we should always free context->module.name, since it will be > allocated unconditionally and audit_log_start() can fail with > other reasons, and audit_log_exit maybe not called > > unreferenced object 0xffff88af90837d20 (size 8): > comm "modprobe", pid 1036, jiffies 4294704867 (age 3069.138s) > hex dump (first 8 bytes): > 69 78 67 62 65 00 ff ff ixgbe... > backtrace: > [<0000000008da28fe>] __audit_log_kern_module+0x33/0x80 > [<00000000c1491e61>] load_module+0x64f/0x3850 > [<000000007fc9ae3f>] __do_sys_init_module+0x218/0x250 > [<0000000000d4a478>] do_syscall_64+0x117/0x400 > [<000000004924ded8>] entry_SYSCALL_64_after_hwframe+0x49/0xbe > [<000000007dc331dd>] 0xffffffffffffffff > > Fixes: ca86cad7380e3 ("audit: log module name on init_module") > Signed-off-by: Zhang Yu <zhangyu31@baidu.com> > Signed-off-by: Li RongQing <lirongqing@baidu.com> > --- > kernel/auditsc.c | 22 ++++++++++++++++++++-- > 1 file changed, 20 insertions(+), 2 deletions(-) > > diff --git a/kernel/auditsc.c b/kernel/auditsc.c > index b2d1f043f..2bd80375f 100644 > --- a/kernel/auditsc.c > +++ b/kernel/auditsc.c > @@ -1186,8 +1186,13 @@ static void show_special(struct audit_context *context, int *call_panic) > int i; > > ab = audit_log_start(context, GFP_KERNEL, context->type); > - if (!ab) > + if (!ab) { > + if (context->type == AUDIT_KERN_MODULE) { > + kfree(context->module.name); > + context->module.name = NULL; > + } > return; > + } Hello. Thanks for the patch, but I have to ask if you've considered freeing the module name in audit_free_context()? That seems like the correct way to solve this issue. -Paul -- paul moore www.paul-moore.com ^ permalink raw reply [flat|nested] 3+ messages in thread
* 答复: [PATCH] audit: fix a memleak caused by auditing load module 2019-03-05 14:18 ` Paul Moore @ 2019-03-06 3:23 ` Li,Rongqing 0 siblings, 0 replies; 3+ messages in thread From: Li,Rongqing @ 2019-03-06 3:23 UTC (permalink / raw) To: Paul Moore; +Cc: Eric Paris, linux-audit, linux-kernel > -----邮件原件----- > 发件人: Paul Moore [mailto:paul@paul-moore.com] > 发送时间: 2019年3月5日 22:18 > 收件人: Li,Rongqing <lirongqing@baidu.com> > 抄送: Eric Paris <eparis@redhat.com>; linux-audit@redhat.com; > linux-kernel@vger.kernel.org > 主题: Re: [PATCH] audit: fix a memleak caused by auditing load module > > On Tue, Mar 5, 2019 at 6:14 AM Li RongQing <lirongqing@baidu.com> wrote: > > we should always free context->module.name, since it will be allocated > > unconditionally and audit_log_start() can fail with other reasons, and > > audit_log_exit maybe not called > > > > unreferenced object 0xffff88af90837d20 (size 8): > > comm "modprobe", pid 1036, jiffies 4294704867 (age 3069.138s) > > hex dump (first 8 bytes): > > 69 78 67 62 65 00 ff ff ixgbe... > > backtrace: > > [<0000000008da28fe>] __audit_log_kern_module+0x33/0x80 > > [<00000000c1491e61>] load_module+0x64f/0x3850 > > [<000000007fc9ae3f>] __do_sys_init_module+0x218/0x250 > > [<0000000000d4a478>] do_syscall_64+0x117/0x400 > > [<000000004924ded8>] entry_SYSCALL_64_after_hwframe+0x49/0xbe > > [<000000007dc331dd>] 0xffffffffffffffff > > > > Fixes: ca86cad7380e3 ("audit: log module name on init_module") > > Signed-off-by: Zhang Yu <zhangyu31@baidu.com> > > Signed-off-by: Li RongQing <lirongqing@baidu.com> > > --- > > kernel/auditsc.c | 22 ++++++++++++++++++++-- > > 1 file changed, 20 insertions(+), 2 deletions(-) > > > > diff --git a/kernel/auditsc.c b/kernel/auditsc.c index > > b2d1f043f..2bd80375f 100644 > > --- a/kernel/auditsc.c > > +++ b/kernel/auditsc.c > > @@ -1186,8 +1186,13 @@ static void show_special(struct audit_context > *context, int *call_panic) > > int i; > > > > ab = audit_log_start(context, GFP_KERNEL, context->type); > > - if (!ab) > > + if (!ab) { > > + if (context->type == AUDIT_KERN_MODULE) { > > + kfree(context->module.name); > > + context->module.name = NULL; > > + } > > return; > > + } > > Hello. > > Thanks for the patch, but I have to ask if you've considered freeing the module > name in audit_free_context()? That seems like the correct way to solve this > issue. > It does not work that move the freeing of module.name in audit_free_context Since we should free module.name based on context->types is AUDIT_KERN_MODULE, but __audit_syscall_exit is called first, and will set context->type to 0, When audit_free_context is called, context->type is 0, will cause to fail. I will change this patches as below: commit ee32ec2354b47a824e5e63d4f46567d577a02824 (HEAD -> master) Author: Li RongQing <lirongqing@baidu.com> Date: Tue Mar 5 15:42:09 2019 +0800 audit: fix a memleak caused by auditing load module module.name will be allocated unconditionally when auditing load module, and audit_log_start() can fail with other reasons, or audit_log_exit maybe not called, caused module.name is released so always free module.name in audit_free_context unreferenced object 0xffff88af90837d20 (size 8): comm "modprobe", pid 1036, jiffies 4294704867 (age 3069.138s) hex dump (first 8 bytes): 69 78 67 62 65 00 ff ff ixgbe... backtrace: [<0000000008da28fe>] __audit_log_kern_module+0x33/0x80 [<00000000c1491e61>] load_module+0x64f/0x3850 [<000000007fc9ae3f>] __do_sys_init_module+0x218/0x250 [<0000000000d4a478>] do_syscall_64+0x117/0x400 [<000000004924ded8>] entry_SYSCALL_64_after_hwframe+0x49/0xbe [<000000007dc331dd>] 0xffffffffffffffff Fixes: ca86cad7380e3 ("audit: log module name on init_module") Signed-off-by: Zhang Yu <zhangyu31@baidu.com> Signed-off-by: Li RongQing <lirongqing@baidu.com> diff --git a/kernel/auditsc.c b/kernel/auditsc.c index b2d1f043f..07728b07a 100644 --- a/kernel/auditsc.c +++ b/kernel/auditsc.c @@ -964,6 +964,9 @@ int audit_alloc(struct task_struct *tsk) static inline void audit_free_context(struct audit_context *context) { + if (context->type == AUDIT_KERN_MODULE) + kfree(context->module.name); + audit_free_names(context); unroll_tree_refs(context, NULL, 0); free_tree_refs(context); @@ -1282,6 +1285,8 @@ static void show_special(struct audit_context *context, int *call_panic) if (context->module.name) { audit_log_untrustedstring(ab, context->module.name); kfree(context->module.name); + context->module.name = NULL; + context->type = 0; } else audit_log_format(ab, "(null)"); @@ -1583,6 +1588,11 @@ void __audit_syscall_exit(int success, long return_code) if (!list_empty(&context->killed_trees)) audit_kill_trees(&context->killed_trees); + if (context->type == AUDIT_KERN_MODULE) { + kfree(context->module.name); + context->module.name = NULL; + } + audit_free_names(context); unroll_tree_refs(context, NULL, 0); audit_free_aux(context); -RongQing > -Paul > > -- > paul moore > www.paul-moore.com ^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-03-06 3:23 UTC | newest] Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-03-05 11:14 [PATCH] audit: fix a memleak caused by auditing load module Li RongQing 2019-03-05 14:18 ` Paul Moore 2019-03-06 3:23 ` 答复: " Li,Rongqing
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).