aio: restore locking of ioctx list on removal
PATCH 1/1 aio: restore locking of ioctx list on removal
Commit 36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted in ioctx_lock not being held during ctx removal, leaving the list susceptible to corruptions.
In mainline kernel the issue went away as a side effect of db446a08c23d5475e6b08c87acca79ebb20f283c "aio: convert the ioctx list to table lookup v3".
Fix the problem by restoring appropriate locking.
--- fs/aio.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
fs/aio.c +6 -4
@@ -423,10 +423,12 @@ static void kill_ioctx_rcu(struct rcu_head *head)* when the processes owning a context have all exited to encourage * the rapid destruction of the kioctx. */ -static void kill_ioctx (struct kioctx *ctx) +static void kill_ioctx (struct mm_struct *mm,struct kioctx *ctx) {if (!atomic_xchg (&ctx->dead,1 )) { +spin_lock (&mm->ioctx_lock);hlist_del_rcu (&ctx->list); +spin_unlock (&mm->ioctx_lock);/* * It'd be more correct to do this in free_ioctx(), after all
@@ -494,7 +496,7 @@ void exit_aio(struct mm_struct *mm)*/ ctx->mmap_size =0 ; -kill_ioctx (ctx); +kill_ioctx (mm, ctx); } }
@@ -852,7 +854,7 @@ SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)if (!IS_ERR (ioctx)) { ret =put_user (ioctx->user_id, ctxp);if (ret) -kill_ioctx (ioctx); +kill_ioctx (current->mm, ioctx);put_ioctx (ioctx); }
@@ -870,7 +872,7 @@ SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx) {struct kioctx *ioctx =lookup_ioctx (ctx);if (likely (NULL != ioctx)) { -kill_ioctx (ioctx); +kill_ioctx (current->mm, ioctx);put_ioctx (ioctx);return 0 ; }
Signature · 1 line
1.8.3.1
Greg KH
maintainerQuoted 9 lines · Mateusz Guzik
Commit
36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted inioctx_locknot being held during ctx removal, leaving the list susceptible to corruptions.In mainline kernel the issue went away as a side effect of
db446a08c23d5475e6b08c87acca79ebb20f283c"aio: convert the ioctx list to table lookup v3".Fix the problem by restoring appropriate locking.
Why can't I just take db446a08c23d5475e6b08c87acca79ebb20f283c instead? Does it not work well enough, or is there other issues involved in it that would keep it out of stable?
Also, it seems like the performance increase of that patch would be good to have backported, right?
thanks,
greg k-h
Quoted 9 lines · Greg KH
Quoted 8 lines · Mateusz Guzik
Commit
36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted inioctx_locknot being held during ctx removal, leaving the list susceptible to corruptions.In mainline kernel the issue went away as a side effect of
db446a08c23d5475e6b08c87acca79ebb20f283c"aio: convert the ioctx list to table lookup v3".
Quoted 9 lines
Quoted 1 line
Fix the problem by restoring appropriate locking.
Why can't I just take
db446a08c23d5475e6b08c87acca79ebb20f283cinstead? Does it not work well enough, or is there other issues involved in it that would keep it out of stable?Also, it seems like the performance increase of that patch would be good to have backported, right?
Sorry, should have noted this in my original message:db446a08c23d5475e6b08c87acca79ebb20f283c is not trivial and applying it results in some conflicts, in addition to that the patch itself had bugs which were fixed in:da90382c2ec367aac88ff6aa76afb659ee0e4235 f30d704fe1244c44a984d3d1f47bc648bcc6c9f7 77d30b14d24e557f89c41980011d72428514d729 d9b2c8714aef102dea95544a8cd9372b21af463f
It may be that the most convienent way to deal with this backport would be to just sync the file with mainline.
As such, I think backporting is too risky at this stage.
Additionally my understanding of Documentation/stable_kernel_rules.txt was that somewhat simpler patches are preferred.
So in the end I decided to fix the problem just by adding locking.
Unfortunately at this time I can't volunteer to do the work if backporting is preferred.
Signature · 1 line
Mateusz Guzik
Hi Greg, Mateusz,
Quoted 29 lines · Mateusz Guzik
Quoted 9 lines · Greg KH
Quoted 8 lines · Mateusz Guzik
Commit
36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted inioctx_locknot being held during ctx removal, leaving the list susceptible to corruptions.In mainline kernel the issue went away as a side effect of
db446a08c23d5475e6b08c87acca79ebb20f283c"aio: convert the ioctx list to table lookup v3".Quoted 9 lines
Quoted 1 line
Fix the problem by restoring appropriate locking.
Why can't I just take
db446a08c23d5475e6b08c87acca79ebb20f283cinstead? Does it not work well enough, or is there other issues involved in it that would keep it out of stable?Also, it seems like the performance increase of that patch would be good to have backported, right?
Sorry, should have noted this in my original message:
db446a08c23d5475e6b08c87acca79ebb20f283cis not trivial and applying it results in some conflicts, in addition to that the patch itself had bugs which were fixed in:da90382c2ec367aac88ff6aa76afb659ee0e4235f30d704fe1244c44a984d3d1f47bc648bcc6c9f777d30b14d24e557f89c41980011d72428514d729d9b2c8714aef102dea95544a8cd9372b21af463f
Quoted 4 lines
It may be that the most convienent way to deal with this backport would be to just sync the file with mainline.
As such, I think backporting is too risky at this stage.
Mateusz's analysis is correct. Backporting db446a08c23d5475e6b08c87acca79ebb20f283c would essentially mean redoing the entire patch and would end up being a bunch more untested code. Please take Mateusz's patch which fixes the bug within a narrow scope. It is applicable to 3.10 and 3.11. For 3.12 the issue is fixed by other changes. Feel free to add my Acked-by.
-ben
Quoted 16 lines
Additionally my understanding of
Documentation/stable_kernel_rules.txtwas that somewhat simpler patches are preferred.So in the end I decided to fix the problem just by adding locking.
Unfortunately at this time I can't volunteer to do the work if backporting is preferred.
Signature · 7 lines
Mateusz Guzik -- To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to [email protected]. For more info on Linux AIO, see: http://www.kvack.org/aio/ Don't email: <a href=mailto:"[email protected]">[email protected]</a>
Signature · 1 line
"Thought is the essence of where you are now."
Quoted 37 lines · Benjamin LaHaise
Hi Greg, Mateusz,
Quoted 29 lines · Mateusz Guzik
Quoted 9 lines · Greg KH
Quoted 8 lines · Mateusz Guzik
Commit
36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted inioctx_locknot being held during ctx removal, leaving the list susceptible to corruptions.In mainline kernel the issue went away as a side effect of
db446a08c23d5475e6b08c87acca79ebb20f283c"aio: convert the ioctx list to table lookup v3".Quoted 9 lines
Quoted 1 line
Fix the problem by restoring appropriate locking.
Why can't I just take
db446a08c23d5475e6b08c87acca79ebb20f283cinstead? Does it not work well enough, or is there other issues involved in it that would keep it out of stable?Also, it seems like the performance increase of that patch would be good to have backported, right?
Sorry, should have noted this in my original message:
db446a08c23d5475e6b08c87acca79ebb20f283cis not trivial and applying it results in some conflicts, in addition to that the patch itself had bugs which were fixed in:da90382c2ec367aac88ff6aa76afb659ee0e4235f30d704fe1244c44a984d3d1f47bc648bcc6c9f777d30b14d24e557f89c41980011d72428514d729d9b2c8714aef102dea95544a8cd9372b21af463fQuoted 4 lines
It may be that the most convienent way to deal with this backport would be to just sync the file with mainline.
As such, I think backporting is too risky at this stage.
Quoted 7 lines
Mateusz's analysis is correct. Backporting
db446a08c23d5475e6b08c87acca79ebb20f283cwould essentially mean redoing the entire patch and would end up being a bunch more untested code. Please take Mateusz's patch which fixes the bug within a narrow scope. It is applicable to 3.10 and 3.11. For 3.12 the issue is fixed by other changes. Feel free to add my Acked-by.-ben
Thank you, I'm queuing it for the 3.11 kernel as well.
Cheers,
--
Luis
Quoted 25 lines
Quoted 18 lines
Additionally my understanding of
Documentation/stable_kernel_rules.txtwas that somewhat simpler patches are preferred.So in the end I decided to fix the problem just by adding locking.
Unfortunately at this time I can't volunteer to do the work if backporting is preferred.
Signature · 7 lines
Mateusz Guzik -- To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to [email protected]. For more info on Linux AIO, see: http://www.kvack.org/aio/ Don't email: <a href=mailto:"[email protected]">[email protected]</a>Signature · 5 lines
"Thought is the essence of where you are now." -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 9 lines · Greg KH
Quoted 8 lines · Mateusz Guzik
Commit
36f5588905c10a8c4568a210d601fe8c3c27e0f0
"aio: refcounting cleanup" resulted inioctx_locknot being held during ctx removal, leaving the list susceptible to corruptions.In mainline kernel the issue went away as a side effect of
db446a08c23d5475e6b08c87acca79ebb20f283c"aio: convert the ioctx list to table lookup v3".
Quoted 8 lines
Quoted 1 line
Fix the problem by restoring appropriate locking.
Why can't I just take
db446a08c23d5475e6b08c87acca79ebb20f283cinstead? Does it not work well enough, or is there other issues involved in it that would keep it out of stable?Also, it seems like the performance increase of that patch would be good to have backported, right?
There are several aio patches that have to be applied to -stable, but I've just been waiting for some feedback now that they're in Linus' tree for a bit before pushing them to -stable. I'll review which ones are needed today and send out a recommended list of patches to apply.
-ben
Quoted 9 lines
thanks,
greg k-h
--
To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to [email protected]. For more info on Linux AIO,
see: http://www.kvack.org/aio/
Don't email: <a href=mailto:"[email protected]">[email protected]</a>
Signature · 1 line
"Thought is the essence of where you are now."