Re: [PATCH 1/2] ipc,shm: move BUG_ON check into shm_lock

[Davidlohr Bueso]

On Thu, 2015-06-04 at 20:35 +0200, Manfred Spraul wrote:
> Hi Davidlohr,
> 
> On 05/30/2015 02:03 AM, Davidlohr Bueso wrote:
> > Upon every shm_lock call, we BUG_ON if an error was returned,
> > indicating racing either in idr or in RMID. Move this logic
> > into the locking.
> >
> > Signed-off-by: Davidlohr Bueso <dbueso@xxxxxxx>
> > ---
> >   ipc/shm.c | 11 +++++++----
> >   1 file changed, 7 insertions(+), 4 deletions(-)
> >
> > diff --git a/ipc/shm.c b/ipc/shm.c
> > index 6d76707..3152dea 100644
> > --- a/ipc/shm.c
> > +++ b/ipc/shm.c
> > @@ -155,8 +155,14 @@ static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
> >   {
> >   	struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
> >   
> > -	if (IS_ERR(ipcp))
> > +	if (IS_ERR(ipcp)) {
> > +		/*
> > +		 * We raced in the idr lookup or with RMID,
> > +		 * either way, the ID is busted.
> The comment is wrong:
> "We raced in the idr lookup or with shm_destroy()"
> 
> shm is not like msg or sem:
> RMID merely marks a shmid as deletable (SHM_DEST), delete (i.e.: 
> shm_rmid(), then ipc_rmid()) happens only after the last attached 
> segment is detatched.
> 
> And: (unrelated to the patch)
> For do_shmat(), I'm not 100% sure that the BUG can't be triggered.

Could you be referring to the fact that we don't mark the segment
->deleted when doing RMID (do_shm_rmid)? This would certainly cause
checks to ipc_valid_object() to completely screw up. I've actually been
chasing that BUG() in shm_open and shm_close during forking and exiting
(unmapping) respectively.

*Completely untested*

Thanks,
Davidlohr

diff --git a/ipc/shm.c b/ipc/shm.c
index 3323c49..4e78cd2 100644
--- a/ipc/shm.c
+++ b/ipc/shm.c
@@ -95,6 +95,14 @@ static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
 		shp->shm_perm.mode |= SHM_DEST;
 		/* Do not find it any more */
 		shp->shm_perm.key = IPC_PRIVATE;
+		/*
+		 * Mark the object deleted. As opposed to
+		 * sems or msg queues, shm only marks the
+		 * shmid deletable upon RMID, and delays
+		 * the freeing until there are no more
+		 * active users of the segment.
+		 */
+		shp->shm_perm.deleted = true;
 		shm_unlock(shp);
 	} else
 		shm_destroy(ns, shp);


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/