Re: Corruption in 2.1.106

Linus Torvalds (torvalds@transmeta.com)
Thu, 18 Jun 1998 08:12:40 -0700 (PDT)

On Fri, 19 Jun 1998, Alan Cox wrote:
>
> Im having fun trying to pin this down for definite as Im using new hardware
> but it definitely appears that 2.1.105 (and + patches) is stablish on my
> hardware. 2.1.106 (with or without patches) is corrupting the file system -
> and normally in ways that dont show up until you force an fsck by hand
> when you find bad blocks, duplicate blocks and interesting crap lurking
> on the wrong parts of the disk.
>
> Are other people seeing 2.1.106 corrupting disks and memory on x86 ?

I haven't gotten any reports per email, but I haven't read the newsgroups
very much as I'm currently in New Orleans for USENIX.

However, 2.1.106 doesn't seem to contain very much new code when it comes
to core functionality. I did the mm_sem thing, but that just _adds_
locking because I was chicken and decided it was imprudent to try to
delete places where the kernel lock was no longer needed.

Looking at the patches, the only other thing that looks suspicious is the
pgd quicklist thing. It plays some games with vmalloc() in particular, but
it looks fine - just something I'd be slightly nervous about.

In fact, I've had a couple of really _positive_ reports about 106 - the
only bad reports I've seen (until yours) looked sound-related. And it
appears you've fixed those with your patches, according to other reports
I've seen..

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu