crashes in clear_user_page

Olaf Hering olh at suse.de
Thu Mar 4 23:47:59 EST 2004


 On Thu, Mar 04, Olaf Hering wrote:

>
>  On Thu, Mar 04, Anton Blanchard wrote:
>
> >
> > Hi,
> >
> > > I get crashes in clear_user_page() while building rpms on a p660.
> > > gcc is 3.2.2, config is arch/ppc64/configs/pseries, plain ameslab tree.
> > >
> > > there is lot of IO, 8 processes do unpack rpms in parallel on a reiserfs
> > > filesystem.
> >
> > It turns out you got a kernel segment with ks set. Shouldnt ever happen.
> >
> > We werent zeroing the old contents of the segment table entry before
> > inserting the new one, if we overwrote a user segment with a kernel one
> > the ks bit would remain set. Its a POWER3/RS64 only bug.
> >
> > Give this patch a go. While I was there I modified our vsid calculation
> > code to match reality (we only use 13 bits of the EA).
>
> does not help, unless we can not blame reiserfs.
                               now
> 5:mon> d c0000000fffeb010
> c0000000fffeb010 **************** ****************  |                |
> 5:mon>

5:mon> u
Segment table contents of cpu 5
000 c000000000000090 00006a99b4b14000
001 e000000000000090 0000a708a8242000
002 d000000000000090 000008d12e6ab000
003 00000080000000b0 0000b6c9b710a000
004 0000008000000030 0000c50de9452000
008 00000000100000b0 00008de66f10a000
024 c000000030000090 0000a12fdcb14000
032 0000000040000030 00005111b3f4b000
064 e000000080000090 00008dee68242000
120 c0000000f00000b0 00007b887cb14000
128 c000000100000090 0000386534b14000
136 c000000110000090 0000f541ecb14000
144 c000000120000090 0000b21ea4b14000
152 c000000130000090 00006efb5cb14000
160 c000000140000090 00002bd814b14000
168 c000000150000090 0000e8b4ccb14000
176 c000000160000090 0000a59184b14000
184 c000000170000090 0000626e3cb14000
248 000001fff00000b0 0000ab2cff10a000


--
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG

** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/





More information about the Linuxppc64-dev mailing list