Re: [PATCH 1/5] block: rewrite blk_bvec_map_sg to avoid a nth_page call

[Guenter Roeck <linux@roeck-us.net>]

On Tue, Apr 16, 2019 at 07:10:11PM +0200, Christoph Hellwig wrote:
> On Tue, Apr 16, 2019 at 10:08:47AM -0700, Guenter Roeck wrote:
> > That is besides the point. Your code changes an internal API to be more
> > stringent and less forgiving. This causes failures, presumably because
> > callers of that API took advantage (on purpose or not) of it.
> > When changing an API, you are responsible for both ends. You can not claim
> > that the callers of that API are buggy. Taking advangage of a forgiving
> > API is not a bug. If you change an API, and that change causes a failure,
> > that is a regression, not a bug on the side of the caller.
> 
> As said I offered to fix these, even if this isn't my fault.  I'm also

"even if this isn't my fault"

Here is where we disagree. You introduced the change, you are responsible
for its impact, on both ends.

> still waiting for the the other reports.

I reported everything I know. To summarize, the following tests are confirmed
to fail due to this patch.

arm:vexpress-a9:multi_v7_defconfig:nolocktests:sd:mem128:vexpress-v2p-ca9:rootfs 
arm:vexpress-a15:multi_v7_defconfig:nolocktests:sd:mem128:vexpress-v2p-ca15-tc1:rootfs 
arm:vexpress-a15-a7:multi_v7_defconfig:nolocktests:sd:mem256:vexpress-v2p-ca15_a7:rootfs 
sparc32:SPARCClassic:nosmp:scsi:hd 
sparc32:SPARCbook:nosmp:scsi:cd 
sparc32:SS-5:nosmp:scsi:hd 
sparc32:SS-10:nosmp:scsi:cd 
sparc32:SS-600MP:nosmp:scsi:hd 
sparc32:Voyager:nosmp:noapc:scsi:hd 
sparc32:SS-4:smp:scsi:hd 
sparc32:SS-5:smp:scsi:cd 
sparc32:SS-20:smp:scsi:hd 
sparc32:SS-600MP:smp:scsi:hd 
sparc32:Voyager:smp:noapc:scsi:hd

Detailed logs are available at https://kerneltests.org/builders, and the
test scripts are published at https://github.com/groeck/linux-build-test.

Guenter