[ltt-dev] [PATCH v2] lttng: Fix potential invalid pointer derefference
Mathieu Desnoyers
compudj at krystal.dyndns.org
Thu Feb 26 16:36:15 EST 2009
* Gui Jianfeng (guijianfeng at cn.fujitsu.com) wrote:
> Mathieu Desnoyers wrote:
> > * Gui Jianfeng (guijianfeng at cn.fujitsu.com) wrote:
> >> Hi Mathieu,
> >>
> >> Calling get_ltt_root() and put_ltt_root() pair may induce invalid pointer
> >> derefference. Consider the following scenario:
> >>
> >> CPU 0 CPU 1
> >> ---------------------- ----------------------
> >> 1 root = get_ltt_root()
> >> 2 root = get_ltt_root()
> >> 3 put_ltt_root() //global root is freed
> >> 4 using root(crash here)
> >>
> >> Here is the fix for this issue. Thanks a lot to Zhaolei to point this out
> >> in offline.
> >>
> >> This patch assumes that the *ltt root remove* patch has been applied.
> >>
> >> Signed-off-by: Gui Jianfeng <guijianfeng at cn.fujistu.com>
> >> ---
> >> include/linux/ltt-core.h | 3 ++-
> >> ltt/ltt-core.c | 25 +++++++++++++++++++++----
> >> ltt/ltt-relay.c | 10 +++++++++-
> >> ltt/ltt-userspace-event.c | 3 +++
> >> 4 files changed, 35 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/include/linux/ltt-core.h b/include/linux/ltt-core.h
> >> index 28394cb..786c2b1 100644
> >> --- a/include/linux/ltt-core.h
> >> +++ b/include/linux/ltt-core.h
> >> @@ -30,9 +30,10 @@ extern struct ltt_traces ltt_traces;
> >> /*
> >> * get dentry of ltt's root dir
> >> */
> >> -void put_ltt_root(void);
> >> struct dentry *get_ltt_root(void);
> >>
> >> +void put_ltt_root(void);
> >> +
> >> /* Keep track of trap nesting inside LTT */
> >> DECLARE_PER_CPU(unsigned int, ltt_nesting);
> >>
> >> diff --git a/ltt/ltt-core.c b/ltt/ltt-core.c
> >> index 314750b..03b81ca 100644
> >> --- a/ltt/ltt-core.c
> >> +++ b/ltt/ltt-core.c
> >> @@ -10,6 +10,7 @@
> >> #include <linux/percpu.h>
> >> #include <linux/module.h>
> >> #include <linux/debugfs.h>
> >> +#include <linux/kref.h>
> >>
> >> /* Traces structures */
> >> struct ltt_traces ltt_traces = {
> >> @@ -23,12 +24,22 @@ static DEFINE_MUTEX(ltt_traces_mutex);
> >>
> >> /* dentry of ltt's root dir */
> >> static struct dentry *ltt_root_dentry;
> >> +static struct kref ltt_root_kref = {
> >> + .refcount = ATOMIC_INIT(0),
> >> +};
> >> +
> >> +static void ltt_root_release(struct kref *ref)
> >> +{
> >> + return;
> >> +}
> >>
> >> void put_ltt_root(void)
> >> {
> >> - if (ltt_root_dentry && list_empty(<t_root_dentry->d_subdirs)) {
> >> - debugfs_remove(ltt_root_dentry);
> >> - ltt_root_dentry = NULL;
> >> + if (ltt_root_dentry) {
> >> + if (kref_put(<t_root_kref, ltt_root_release)) {
> >> + debugfs_remove(ltt_root_dentry);
> >> + ltt_root_dentry = NULL;
> >
> > Those two should go in ltt_root_release. See other (wide) use of
> > kref_get/kref_put in the LTT code as example.
> >
> > Mathieu
> >
> >> + }
> >> }
> >> }
> >> EXPORT_SYMBOL_GPL(put_ltt_root);
> >> @@ -37,9 +48,15 @@ struct dentry *get_ltt_root(void)
> >> {
> >> if (!ltt_root_dentry) {
> >> ltt_root_dentry = debugfs_create_dir(LTT_ROOT, NULL);
> >> - if (!ltt_root_dentry)
> >> + if (!ltt_root_dentry) {
> >> printk(KERN_ERR "LTT : create ltt root dir failed\n");
> >> + goto out;
> >> + }
> >> + kref_init(<t_root_kref);
> >> + goto out;
> >> }
> >> + kref_get(<t_root_kref);
> >> +out:
> >> return ltt_root_dentry;
> >> }
> >> EXPORT_SYMBOL_GPL(get_ltt_root);
> >> diff --git a/ltt/ltt-relay.c b/ltt/ltt-relay.c
> >> index 7c544f4..8378802 100644
> >> --- a/ltt/ltt-relay.c
> >> +++ b/ltt/ltt-relay.c
> >> @@ -833,8 +833,16 @@ end:
> >>
> >> static int ltt_relay_create_dirs(struct ltt_trace_struct *new_trace)
> >> {
> >> + struct dentry *ltt_root_dentry;
> >> +
> >> + ltt_root_dentry = get_ltt_root();
> >> + if (!ltt_root_dentry)
> >> + return ENOENT;
> >> +
> >> new_trace->dentry.trace_root = debugfs_create_dir(new_trace->trace_name,
> >> - get_ltt_root());
> >> + ltt_root_dentry);
> >> + put_ltt_root();
> >> + ltt_root_dentry = NULL;
> >
> > If ltt_root_dentry is local to the function, I don't see real value in
> > setting it to null ?
>
> Ok, will remove.
>
> >
> > You should also modifify ltt-relay-locked, because it has almost the
> > exact same code as ltt-relay.c.
>
> ltt-relay-locked doesn't use get_ltt_root(), it has its own root directory
> called "ltt-locked", so i don't see why ltt-relay-locked should be modified.
>
Yes, you are right.
Mathieu
> >
> > Thanks,
> >
> > Mathieu
> >
> >> if (new_trace->dentry.trace_root == NULL) {
> >> printk(KERN_ERR "LTT : Trace directory name %s already taken\n",
> >> new_trace->trace_name);
> >> diff --git a/ltt/ltt-userspace-event.c b/ltt/ltt-userspace-event.c
> >> index a48f336..487516b 100644
> >> --- a/ltt/ltt-userspace-event.c
> >> +++ b/ltt/ltt-userspace-event.c
> >> @@ -110,7 +110,10 @@ static int __init ltt_userspace_init(void)
> >> err = -EPERM;
> >> goto err_no_file;
> >> }
> >> +
> >> + return err;
> >> err_no_file:
> >> + put_ltt_root();
> >> err_no_root:
> >> return err;
> >> }
> >> --
> >> 1.5.4.rc3
> >>
> >>
> >> _______________________________________________
> >> ltt-dev mailing list
> >> ltt-dev at lists.casi.polymtl.ca
> >> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
> >>
> >
>
> --
> Regards
> Gui Jianfeng
>
--
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
More information about the lttng-dev
mailing list