On Sun, Dec 7, 2008 at 11:50 PM, John David Anglin <dave@xxxxxxxxxxxxxxxxxx> wrote: >> Given your earlier comments it would seen the segfault is not >> reproducible and happens often and randomly? > > I gave 2.6.28-rc7-git5 a try this evening. Hit the same error. > While its occurence may be random, the identical fault occurred > multiple times in a short period of testing. > > I see this bit of code in _dl_fixup: > > 0x401ed464 <_dl_fixup+172>: mfctl tr3,r21 > 0x401ed468 <_dl_fixup+176>: ldw -3e0(r21),ret0 > 0x401ed46c <_dl_fixup+180>: cmpib,<> 0,ret0,0x401ed628 <_dl_fixup+624> > > and > > 0x401ed5fc <_dl_fixup+580>: ldi 1,r24 > 0x401ed600 <_dl_fixup+584>: ldi 0,r23 > 0x401ed604 <_dl_fixup+588>: mfctl tr3,r26 > 0x401ed608 <_dl_fixup+592>: copy r24,r25 > 0x401ed60c <_dl_fixup+596>: ldo -3dc(r26),r26 > 0x401ed610 <_dl_fixup+600>: copy r19,r4 > 0x401ed614 <_dl_fixup+604>: be,l 100(sr2,r0),sr0,r31 > 0x401ed618 <_dl_fixup+608>: ldi d2,r20 > > So, the code is using the thread register (cr27). Not sure what syscall > this is. However, the problem might be TLS releated. The thread register has always been used to track thread specific information, even in linuxthreads. The syscall is __NR_futex. If you can produce a synthetic testcase that calls __NR_futex and crashes the kernel then we'll have something for the kernel hackers to look at? Cheers, Carlos. -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
