Re: [PATCH v1 0/2] spi: Make dummy SG handling robust

[Nícolas F. R. A. Prado <nfraprado@xxxxxxxxxxxxx>]

On Sat, Jun 01, 2024 at 12:45:30AM +0300, Andy Shevchenko wrote:
> On Fri, May 31, 2024 at 8:14 PM Nícolas F. R. A. Prado
> <nfraprado@xxxxxxxxxxxxx> wrote:
> >
> > On Fri, May 31, 2024 at 06:51:46PM +0300, Andy Shevchenko wrote:
> > > On Fri, May 31, 2024 at 6:46 PM Andy Shevchenko
> > > <andy.shevchenko@xxxxxxxxx> wrote:
> > > > On Fri, May 31, 2024 at 5:37 PM Nícolas F. R. A. Prado
> > > > <nfraprado@xxxxxxxxxxxxx> wrote:
> > > > > On Fri, May 31, 2024 at 12:44:31PM +0300, Andy Shevchenko wrote:
> 
> ...
> 
> > > > > applying either of these patches causes issues. See the traces for each one
> > > > > below. This was tested on top of next-20240531, which works fine.
> > > >
> > > > Oh, thank you very much for prompt testing! Can you test just the
> > > > second one without the revert?
> > >
> > > Ah, you wrote "either", so it seems you have tried that already.
> >
> > Yes exactly. Both patches are troublesome. Patch 2 causes a slightly different
> > null pointer dereference, in "dcache_clean_poc+0x20/0x38", as the stack trace I
> > posted shows.
> 
> I have sent a new series where the last patch has a massive rework of
> the cur_msg_mapped flag. Would be nice to see if it passes your tests.
> The main idea there is to actually move to per transfer flag(s) from
> per message one.

Ah great, I really felt that the flag should've been per transfer, so thank you
for making that change. I'll do some testing on Monday.

Thanks,
Nícolas