Hi Lars, > -----Original Message----- > From: Lars-Peter Clausen [mailto:[email protected]] > Sent: Monday, February 06, 2017 9:40 PM > > On 02/06/2017 01:30 PM, Yoshihiro Shimoda wrote: > > From: Hiroyuki Yokoyama <[email protected]> > > > > This patch fixes the problem that occasionally released resources > > before the end of interrupt processing. > > > > [ 58.156412] Unable to handle kernel NULL pointer dereference at virtual > > address 00000000 > > [ 58.166155] pgd = ffff8006f78b0000 > > [ 58.169822] [00000000] *pgd=000000073773b003 , *pud=0000000737c70003 , > > *pmd=0000000000000000 > > [ 58.179738] > > [ 58.181548] Internal error: Oops: 96000046 [#1] PREEMPT SMP > > [ 58.187484] Modules linked in: > > [ 58.190919] CPU: 0 PID: 2898 Comm: dma_ioctl Not tainted 4.9.0-00002-g501 > > 07f2-dirty #147 > > [ 58.199438] Hardware name: Renesas Salvator-X board based on r8a7795 (DT) > > [ 58.206680] task: ffff8006f77b1900 task.stack: ffff8006f7bd4000 > > [ 58.213090] PC is at rcar_dmac_chan_prep_sg+0xa4/0x3f0 > > [ 58.218725] LR is at rcar_dmac_chan_prep_sg+0x6c/0x3f0 > > > > Signed-off-by: Hiroyuki Yokoyama <[email protected]> > > Signed-off-by: Yoshihiro Shimoda <[email protected]> > > How about implementing the synchronize() callback. This will fix not only > this race condition, but also others that can result from the same issue. > E.g. the completion callback still running after terminate() has completed.
Thank you for the suggestion! I will investigate how to implement device_snchronize() in this driver. Best regards, Yoshihiro Shimoda
