Re: [BUG] "block: make generic_make_request handle arbitrarily sized bios" breaks boot on parisc-linux

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, 2016-03-09 at 22:20 +0100, Helge Deller wrote:
> On 09.03.2016 16:15, John David Anglin wrote:
> > On 2016-03-09 9:43 AM, Ming Lei wrote:
> > > > We've provided all the information you asked for, what's the
> > > > next step
> > > > > on this, or do we have to unwind the bio splitting code with
> > > > > reverts
> > > > > until it starts working?
> > > John, Helge, and I did discuss the problem for a while privately,
> > > and looks
> > > it is related with compiler.  Last time, I sent one patch which
> > > can make the
> > > issue disappear, but the main change is just invovled with the
> > > below:
> > > 
> > >   struct bio_vec {
> > >       struct page    *bv_page;
> > > -    unsigned int    bv_len;
> > > +    unsigned int    bv_seg:8;
> > > +    unsigned int    bv_len:24;
> > >       unsigned int    bv_offset;
> > >   };
> > > 
> > > Maybe John and Helge have some update recently?
> > > 
> > > The logic in blk_bio_segment_split() is correct, and it does
> > > respect the max
> > > segment size limit.
> > Helge has found that tagging blk_bio_segment_split() with
> > "__attribute__ ((optimize("O0")))"
> > makes the issue disappear.  The bug remains if one just adds bv_len
> > to the struct without the
> > bit fields.  Maybe problem is evident from following output which I
> > sent to Ming and Helge
> > last weekend?
> > 
> > blk_rq_map_sg: merge bug: 3 2, extra_len 0, dma_drain 0
> > check_bvec: dump bvec for 000000007e4efdc0(f:24490000, t:1)
> >             0: 0 4096 246503 000000007e4a4f00(0, 94208, 1)
> >             1: 0 4096 246504 000000007e4a4f00(0, 94208, 1)
> >             2: 0 4096 246505 000000007e4a4f00(0, 94208, 1)
> >             3: 0 4096 246506 000000007e4a4f00(0, 94208, 1)
> >             4: 0 4096 246538 000000007e4a4f00(0, 94208, 2)
> >             5: 0 4096 246539 000000007e4a4f00(0, 94208, 2)
> >             6: 0 4096 246540 000000007e4a4f00(0, 94208, 2)
> >             7: 0 4096 246541 000000007e4a4f00(0, 94208, 2)
> >             8: 0 4096 246542 000000007e4a4f00(0, 94208, 2)
> >             9: 0 4096 246543 000000007e4a4f00(0, 94208, 2)
> >            10: 0 4096 246544 000000007e4a4f00(0, 94208, 2)
> >            11: 0 4096 246545 000000007e4a4f00(0, 94208, 2)
> >            12: 0 4096 246546 000000007e4a4f00(0, 94208, 2)
> >            13: 0 4096 246547 000000007e4a4f00(0, 94208, 2)
> >            14: 0 4096 246548 000000007e4a4f00(0, 94208, 2)
> >            15: 0 4096 246549 000000007e4a4f00(0, 94208, 2)
> >            16: 0 4096 246550 000000007e4a4f00(0, 94208, 2)
> >            17: 0 4096 246551 000000007e4a4f00(0, 94208, 2)
> >            18: 0 4096 246552 000000007e4a4f00(0, 94208, 2)
> >            19: 0 4096 246553 000000007e4a4f00(0, 94208, 2)
> >            20: 0 4096 246554 000000007e4a4f00(0, 94208, 2)
> >            21: 0 4096 246555 000000007e4a4f00(0, 94208, 2)
> >            22: 0 4096 246556 000000007e4a4f00(0, 94208, 2)
> > Kernel panic - not syncing: bad block merge
> > 
> > It seems segment 1 is too small and segment 2 too big?
> > 
> > The general plan is to disable inlining (maybe move
> > blk_bio_segment_split() to a separate
> > function) to try to figure out what is miscompiled.
> 
> Right.
> I just succeeded in reproducing the bug with moving
> blk_bio_segment_split() into an own file
> (and with "extern" instead of "static" in blk-merge.c). When compiled
> with -O2 it still crashes.
> So, next step is to analyze what gcc does wrong when compiling this
> function.
> It should get easier now to find the reason, since we have a smaller 
> reproducer now.

OK, that would explain why I don't see the problem, since I'm using an
older compiler.  So it's our issue and basically no action for block.

James


--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [SCSI Target Devel]     [Linux SCSI Target Infrastructure]     [Kernel Newbies]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Linux IIO]     [Samba]     [Device Mapper]
  Powered by Linux