Hi Tim, I''m running OLTP on x86_64 SMP FC5 HVM guest, but I saw this messages on guest console: mysqld: Corrupted page table at address 2aaaae717000 PGD 33ae0067 PUD 33ae1067 PMD 2ea4c067 PTE 0 Bad pagetable: 000f [1] SMP last sysfs file: /block/hdc/size CPU 1 Modules linked in: ipv6 autofs4 hidp l2cap bluetooth sunrpc dm_mirror dm_mod video button battery ac lp parport_pc parport floppy nvram 8139cp 8139too mii ext3 jbd Pid: 1966, comm: mysqld Not tainted 2.6.15-1.2054_FC5 #1 RIP: 0033:[<000000000081d76c>] [<000000000081d76c>] RSP: 002b:0000000045108cb0 EFLAGS: 00010206 RAX: 0000000000000020 RBX: 0000000000000078 RCX: 000000000154d529 RDX: 0000000000000014 RSI: 0000000000000078 RDI: 0000000000b7ce40 RBP: 00000000013e0580 R08: 0000000000000078 R09: 00002aaaae717000 R10: 0000000005991432 R11: 000000000081d730 R12: 00002aaaae716fed R13: 0000000000002061 R14: 0000000001605540 R15: 00000000000f424b FS: 000000004510b950(0063) GS:ffff8100021f60c0(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00002aaaae717000 CR3: 0000000034f8b000 CR4: 00000000000006e0 Process mysqld (pid: 1966, threadinfo ffff81002c464000, task ffff81003263b100) RIP [<000000000081d76c>] RSP <0000000045108cb0> Not sure if it''s really a shadow bug, but it''s not hard to reproduce: 1) install FC5 with all package, so mysql will be installed; 2) download and install sysbench 0.4.8; 3) run the script attached to start oltp. After some while, you may saw the warning, if you can not reproduce it, pls tell me. Thanks -Xin _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hi, At 16:49 +0800 on 14 Nov (1163522982), Li, Xin B wrote:> Hi Tim, > I''m running OLTP on x86_64 SMP FC5 HVM guest, but I saw this messages on > guest console: > > mysqld: Corrupted page table at address 2aaaae717000 > PGD 33ae0067 PUD 33ae1067 PMD 2ea4c067 PTE 0 > Bad pagetable: 000f [1] SMPAh; it looks like x86_64 linux doesn''t handle spurious reserved-bit page faults. Does this patch fix things for you? diff -r 032578b83641 xen/arch/x86/mm/shadow/multi.c --- a/xen/arch/x86/mm/shadow/multi.c Tue Nov 14 10:39:49 2006 +0000 +++ b/xen/arch/x86/mm/shadow/multi.c Tue Nov 14 10:43:08 2006 +0000 @@ -2623,6 +2623,10 @@ static int sh_page_fault(struct vcpu *v, * Fall through to the normal fault handing logic */ perfc_incrc(shadow_fault_fast_fail); SHADOW_PRINTK("fast path false alarm!\n"); + /* Don''t pass the reserved-bit bit: if we look at the fault + * below and decide to pass it to the guest, the reserved-bit + * bit won''t make sense there. */ + regs->error_code &= ~PFEC_reserved_bit; } } #endif /* SHOPT_FAST_FAULT_PATH */ Cheers, Tim. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> >At 16:49 +0800 on 14 Nov (1163522982), Li, Xin B wrote: >> Hi Tim, >> I''m running OLTP on x86_64 SMP FC5 HVM guest, but I saw this >messages on >> guest console: >> >> mysqld: Corrupted page table at address 2aaaae717000 >> PGD 33ae0067 PUD 33ae1067 PMD 2ea4c067 PTE 0 >> Bad pagetable: 000f [1] SMP > >Ah; it looks like x86_64 linux doesn''t handle spurious >reserved-bit page >faults. Does this patch fix things for you? > >diff -r 032578b83641 xen/arch/x86/mm/shadow/multi.c >--- a/xen/arch/x86/mm/shadow/multi.c Tue Nov 14 10:39:49 2006 +0000 >+++ b/xen/arch/x86/mm/shadow/multi.c Tue Nov 14 10:43:08 2006 +0000 >@@ -2623,6 +2623,10 @@ static int sh_page_fault(struct vcpu *v, > * Fall through to the normal fault handing logic */ > perfc_incrc(shadow_fault_fast_fail); > SHADOW_PRINTK("fast path false alarm!\n"); >+ /* Don''t pass the reserved-bit bit: if we look at >the fault >+ * below and decide to pass it to the guest, the >reserved-bit >+ * bit won''t make sense there. */ >+ regs->error_code &= ~PFEC_reserved_bit; > } > } > #endif /* SHOPT_FAST_FAULT_PATH */ > >Yes, it''s fixed. thanks -Xin _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel