Re: Linux 2.2.15pre6

[TenThumbs]

Alan Cox wrote:
> 
> 2.2.15pre6
> o       SMC-mca updates                                 (David Monro)
> o       Stallion driver update                          (Wayne Meissner)
> o       DAC960 updates                                  (Leonard Zubkoff)
> o       CPQ Array updates (adds SA431)                  (Charles White)
> o       Fix creative joystick crash
> o       Fix bogus RST when fins crossed                 (Dave Miller)
> o       Gemini serial driver update                     (Cort Dougan)
> o       Work around some MediaGX DMA bugs               (me)
>         - Info thanks to Nat Semi
> o       Fix type clashes between irda and Ingo HSM      (Dag Brattli)
> o       IDE DMA abort handling fixes                    (Andre Hedrick)
> o       Cyrix 5530 IDE support                          (Mark Lord)
> o       3Ware raid controller driver                    (Adam Radford)
> o       Fix sunrpc queue corruption bug                 (Trond Myklebust)
> o       Sparc config updates                            (Dave Miller)
> o       Fix SMP powermac booting                        (Paul Mackerras)
> o       Fix bonding documentation                       (Lennie Besselink)
> o       CPIA parallel camera driver                     (Peter Pregler)
> o       Alternative page allocation hang fixes          (Andrea Arcangeli)
>         | Reverted some of the other fixes. This wants
>         | sorting nicely for 2.3.x, for 2.2.x Andrea's
>         | approach is easier to verify
> o       Fix shm/remap bug                               (Eric Biederman)
> 

I think part of the boobytrap patch is missing. I've been running
pre5+boobytrap2 without any warnings but the first time I tried a fetchmail
run I got all these.

      2 gfp called by non-running (1) task from c004c000!
      3 gfp called by non-running (1) task from c0052000!
      1 gfp called by non-running (1) task from c02ce000!
      2 gfp called by non-running (1) task from c0398000!
      2 gfp called by non-running (1) task from c03d8000!
      1 gfp called by non-running (1) task from c0ed2000!
      1 gfp called by non-running (1) task from c0f30000!
      1 gfp called by non-running (1) task from c0f84000!
      1 gfp called by non-running (1) task from c13a4040!
      1 gfp called by non-running (1) task from c13a4680!
     17 gfp called by non-running (1) task from c1864000!
      3 gfp called by non-running (1) task from c1898000!
      1 gfp called by non-running (1) task from c2142000!

Every address seems to be bogus, too. They're way out of range for the kernel.
Kernel symbols run from c0106000 to c01e964c and the modules start at
c3045000.

I built with gcc 2.7.2.3 so it's possible it has a problem with
__builtin_return_address but I doubt it.

--
A sweet spot promotes, and the progressive constraints agree to
disagree on an effective write-off.