]> git.hungrycats.org Git - linux/commit
dmaengine: edma: fix memory leak when terminating running transfers
authorPetr Kulhavy <petr@barix.com>
Fri, 27 Mar 2015 11:35:51 +0000 (13:35 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 19 Apr 2015 08:10:20 +0000 (10:10 +0200)
commit56e1c4b3236721b26f059d0f331e2464a5d68fd7
treee11658f2294981d8b25a9f2bad25e47cedae0b9e
parent7f754cc8edfb11d255eb91c823f93aa09ef0f138
dmaengine: edma: fix memory leak when terminating running transfers

commit 5ca9e7ce6eebec53362ff779264143860ccf68cd upstream.

If edma_terminate_all() was called while a transfer was running (i.e. after
edma_execute() but before edma_callback()) the echan->edesc was not freed.

This was due to the fact that a running transfer is on none of the
vchan lists: desc_submitted, desc_issued, desc_completed (edma_execute()
removes it from the desc_issued list), so the vchan_dma_desc_free_list()
called at the end of edma_terminate_all() didn't find it and didn't free it.

This bug was found on an AM1808 based hardware (very similar to da850evm,
however using the second MMC/SD controller), where intense operations on the SD
card wasted the device 128MB RAM within a couple of days.

Peter Ujfalusi:
The issue is even more severe since it affects cyclic (audio) transfers as
well. In this case starting/stopping audio will results memory leak.

Signed-off-by: Petr Kulhavy <petr@barix.com>
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
CC: <linux-omap@vger.kernel.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/dma/edma.c