Re: bpf selftest failed in 5.4.210 kernel

[Jean-Philippe Brucker]

On Thu, Aug 18, 2022 at 07:24:03AM +0200, Greg KH wrote:
> On Wed, Aug 17, 2022 at 09:22:00PM +0300, RAJESH DASARI wrote:
> > Hi ,
> > 
> > We are running bpf selftests on 5.4.210 kernel version and we see that
> > test case 11 of  test_align failed. Please find the below error.
> > 
> > selftests: bpf: test_align
> > Test  11: pointer variable subtraction ... Failed to find match 16:
> > R5_w=pkt(id=2,off=0,r=0,umin_value=2,umax_value=1082,var_off=(0x2;
> > 0xfffffffc)
> > # func#0 @0
> > # 0: R1=ctx(id=0,off=0,imm=0) R10=fp0
> > # 0: (61) r2 = *(u32 *)(r1 +76)
> > # 1: R1=ctx(id=0,off=0,imm=0) R2_w=pkt(id=0,off=0,r=0,imm=0) R10=fp0
> > # 1: (61) r3 = *(u32 *)(r1 +80)
> > 
> > For complete errors please see the attached file. The same test case
> > execution was successful in the 5.4.209 version , could you please let
> > me know any known issue with the recent changes in 5.4.210 and how to
> > fix these errors.
> 
> Can you use 'git bisect' to find the offending commit?

It probably is 6098562ed9df ("selftests/bpf: Fix "dubious pointer
arithmetic" test")
https://lore.kernel.org/all/20220803145005.2385039-6-ovidiu.panait@xxxxxxxxxxxxx/
Could you try reverting that?

The patch didn't have a Fixes: tags, because the bugfix it refers to was
merged at the same time. That bugfix is upstream commit b02709587ea3
("bpf: Fix propagation of 32-bit signed bounds from 64-bit bounds.")

Since b02709587ea3 was only backported down to 5.10, this fix shouldn't be
in 5.4. Sorry for not catching this earlier.

Thanks,
Jean