On 03/03/2011 06:01 AM, Yasunori Goto wrote:
In this log, cpu4 and 6 repeat page faults. ---- handle_mm_fault jiffies64=4295160616 cpu=4 address=40019a38 pmdval=0000000070832067 ptehigh=00000000 ptelow=55171067 handle_mm_fault jiffies64=4295160616 cpu=6 address=40003a38 pmdval=0000000070832067 ptehigh=00000000 ptelow=551ef067 handle_mm_fault jiffies64=4295160616 cpu=6 address=40003a38 pmdval=0000000070832067 ptehigh=00000000 ptelow=551ef067 handle_mm_fault jiffies64=4295160616 cpu=4 address=40019a38 pmdval=0000000070832067 ptehigh=00000000 ptelow=55171067 handle_mm_fault jiffies64=4295160616 cpu=4 address=40019a38 pmdval=0000000070832067 ptehigh=00000000 ptelow=55171067
I confirmed this phenomenon is reproduced on 2.6.31 and 2.6.38-rc5 of x86 kernel, and I heard this phenomenon doesn't occur on x86-64 kernel from another engineer who found this problem first. In addition, this phenomenon occurred on 4 boxes, so I think the cause is not hardware malfunction.
On what CPU model(s) does this happen? Obviously the PTE is present and allows read, write and execute accesses, so the PTE should not cause any faults. That leaves the TLB. It looks almost like the CPU keeps re-faulting on a (old?) TLB entry, possibly with wrong permissions, and does not re-load it from the PTE. I know this "should not happen" on x86, but I cannot think of an alternative explanation right now. Can you try flushing the TLB entry in question from handle_pte_fault? It looks like the code already does this for write faults, but maybe the garbage collection code uses PROT_NONE a lot and is running into this issue with a read or exec fault? It would be good to print the fault flags as well in your debug print, so we know what kind of fault is being repeated... -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>