On Wed, Jul 21, 2004 at 11:13:30PM -0700, Joshua Schmidlkofer wrote:
> FWIW I am getting this error. I assumed it was my NVIDIA binary
> drivers causing it. However, there are lots of weird allocator
> issues popping up right now [so it seems].
i don't think it's an allocator problem, see:
http://oss.sgi.com/bugzilla/show_bug.cgi?id=311
> I have checked my FS several times, and it does not seem to be a
> problem. It only happens when I have large blocks of memory sucked
> into some process, which would lead me to believe that there is some
> amount of memory pressure leading up to them.
what kernel? this was fixed May 27 if it's the above problem.
> AFAICT the XFS folks tend to "dead-lock instead of corrupt" [which
> is standard fare in the Linux kernel].
linux memory allocations can fail, linux has always been this way and
linux drivers are written to deal with this
irix memory allocations never fail, so XFS isn't written to deal with
this and thus there is a layer for such cases where XFS tries really
hard to get memory and loops
as best as i can tell current xfs shouldn't explode on low memory but
might stall --- i'm not sure there is much that can be done about that
easily though
i really think most of the pain people are presently seeing will go
away when the CVS tree is sync'd to mainline (or just use CVS). then
you can find a whole lot more pain to fix.
--cw
|