[PATCH 0/2] dma-mapping, remoteproc: Fix dma_mem leak after rproc_shutdown

Mark-PK Tsai mark-pk.tsai at mediatek.com
Mon May 23 05:27:23 PDT 2022


> >> Sigh.  In theory drivers should never declare coherent memory like
> >> this, and there has been some work to fix remoteproc in that regard.
> >>
> >> But I guess until that is merged we'll need somthing like this fix.
> > 
> > Hi,
> > 
> > Thanks for your comment.
> > As I didn't see other fix of this issue, should we use this patch
> > before the remoteproc work you mentioned is merged?
> 
> TBH I think it would be better "fixed" with a kmemleak_ignore() and a 
> big comment, rather than adding API cruft for something that isn't a 
> real problem. I'm quite sure that no real-world user is unbinding 
> remoteproc drivers frequently enough that leaking a 48-byte allocation 
> each time has any practical significance.
> 

Actually we stop 2 remote processors using the same remoteproc driver
when system suspend or screen off on our arm64 platform.
And the 48-byte slab allocation will be aligned up to 128 bytes on arm64.
So the leak can be significant in our use case especially
in stress test...

We really don't want to ignore it. Maybe there're other way to fix
it without adding APIs?



More information about the linux-arm-kernel mailing list