possibly dead lock

[zhizhouzh]

Hi experts,

I met a dead lock issue, the log is shown as below:

D/TC:1 pgt_alloc:527 Waiting for page tables
D/TC:1 __wq_rpc:38 sleep thread 1 0x100069158 -2 core/arch/arm/mm/pgt_cache.c:529
D/TC:0 __wq_rpc:38 sleep thread 0 0x1000695a8 1 core/kernel/tee_ta_manager.c:216
D/TC:2 tee_ta_close_session:403 tee_ta_close_session(0x8fb40)
D/TC:2 __wq_rpc:38 sleep thread 2 0x1000695a8 1 core/kernel/tee_ta_manager.c:216

It seems, one thread holds &tee_ta_mutex and tries to seize &pgt_mu by condvar_wait(&pgt_cv, &pgt_mu);. It's probably because of there is no enough pgt_cache entries, then this thread hangs itself to wait other thread to exit and to free pgt.

But on the other side, the other thread is waiting for &tee_ta_mutex in tee_ta_close_session(). But this lock is seized by the first thread. So this thread is not able to free pgt. Finally results dead lock.

Could anyone help to confirm it? Thanks!

[jenswi-linaro]

Thanks @zhizhouzh, your analysis is correct as far as I can tell. I'll try to fix this.

[jenswi-linaro]

Please test #2083 and tell if it fixes the deadlock.

[zhizhouzh]

Thanks, This issue does not reproduce any more with this patchset.

[jforissier]

Re-opening since the fix (#2083) causes stability issues, and was reverted (#2103).

[zhizhouzh]

Hi, any updates about this issue? Thanks!

[jenswi-linaro]

Unfortunately no.

[github-actions]

This issue has been marked as a stale issue because it has been open (more than) 30 days with no activity. Remove the stale label or add a comment saying that you would like to have the label removed otherwise this issue will automatically be closed in 5 days. Note, that you can always re-open a closed issue at any time.