LWN: Comments on "Kernel release status"

Kernel release status

nobrowser — Fri, 07 May 2004 17:49:49 +0000

What happened to the ext[23] htree feature? It was promised for 2.4.26 but
it's not there; and now 2.4 seems to be nearing death.

2.6.6-rc3 crashes, here -- FIXED

Duncan — Sat, 01 May 2004 16:52:30 +0000

Seems that mine has been fixed with the 2.6.6-rc3-bk3, which I'm now
running. I expected it was, when I saw this in the log:

<quote>

ChangeSet@1.1638, 2004-04-30 14:51:54-07:00, akpm@osdl.org
[PATCH] task_struct alignment fix

The recent slab alignment changes broke [many archs] (parisc and x86_64
for sure) by causing task_structs to be insufficiently aligned. [C]hange
[it back] to L1_CACHE_BYTES, [as] we used to have, via
SLAB_HW_CACHE_ALIGN.

</quote>

I figured that must be what I saw on x86_64, but I can boot the new kernel
again, now, and as a bonus I now know what the nightly bk snapshots are
(as mentioned in /something/ I was reading, recently, either on the Gentoo
lists or on LWN) and where to find and how to apply them. =:^)

Duncan

2.6.6-rc3 crashes, here.

Duncan — Fri, 30 Apr 2004 17:09:40 +0000

I don't think it's the same bug, then, as I'm running rc2 without issue.

Unless it was in rc1 (which I didn't have time to try), somehow got masked
at least for my combination of hardware and kernel config for rc2, and
then reappeared in rc3..

I suppose that's possible, if it's something under current rapid
development. However, I wouldn't think it all that likely..

Maybe I'll have to go pull rc1, compile it, and see, if 2.6.6 doesn't fix
it or is delayed until after the weekend. Right now, I have the sleep I
SHOULD have been doing when I was working on the computer, to catch up on.
<g>

Duncan

2.6.6-rc3 crashes, here.

corbet — Fri, 30 Apr 2004 14:21:39 +0000

FWIW, I believe I'm begin bitten by the same bug; it was introduced in -rc1. I'm trying to track it down in my spare time...

2.6.6-rc3 crashes, here.

Duncan — Fri, 30 Apr 2004 14:12:27 +0000

2.6.6-rc3 crashes, here. It compiles fine, but it crashes either at /proc
mount or shortly thereafter, giving me an error about sleeping with a lock
held, and preemption.

I'm running dual Opteron 242s, thus AMD64 in SMP mode. The problem quite
likely is some 64-bit unclean code somewhere, very possibly not triggered
and possibly compiled out, unless SMP and/or preemption are on.

Unfortunately, I don't know enough about kernel bug reporting to know how
or what to do about it, or even how to check to see if anyone else has
reported something similar, so it's just wait and see if it's fixed in
2.6.6 proper, right now. If not, I'll have to go read up on all that and
figure it out. (I may have done so already, except the hard crash took
out my video card and it took me several hours to figure out exactly what
it was and fix it.. in part because the card would come back after some
twiddling around for awhile, as if certain crashes stuffed up a
configuration register that took some time to stabilize due to capacitance
bleed-off, but the card now seems to have quit for good.. it's not coming
back any more.. Oh, well, it was only an ATI Radeon 7200 dual monitor
out, so not TOO expensive.)

Duncan