| Message ID | 20100308120009.GC14199@in.ibm.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
Ananth N Mavinakayanahalli wrote: >>> 2.6.33-git10(64ba99267...) was OK. >>> >>> This WARN_ON was introduced by commit 4610ee1d36... >>> >>> kprobes: Introduce generic insn_slot framework >>> >> FWIW, same on s390... >> > > This patch should fix it: > > When freeing the instruction slot, the arithmetic to calculate the index > of the slot in the page needs to account for the total size of the > instruction on the various architectures. > > Calculate the index correctly when freeing the out-of-line execution slot. > > Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com> > Thanks Ananth. Boots fine on my POWER6 box without any WAR_ON. Regards -Sachin > --- > Index: linux-8mar/kernel/kprobes.c > =================================================================== > --- linux-8mar.orig/kernel/kprobes.c 2010-03-08 17:10:33.000000000 +0530 > +++ linux-8mar/kernel/kprobes.c 2010-03-08 17:12:12.000000000 +0530 > @@ -259,7 +259,8 @@ > struct kprobe_insn_page *kip; > > list_for_each_entry(kip, &c->pages, list) { > - long idx = ((long)slot - (long)kip->insns) / c->insn_size; > + long idx = ((long)slot - (long)kip->insns) / > + (c->insn_size * sizeof(kprobe_opcode_t)); > if (idx >= 0 && idx < slots_per_page(c)) { > WARN_ON(kip->slot_used[idx] != SLOT_USED); > if (dirty) { > > > >
Ananth N Mavinakayanahalli wrote: > On Mon, Mar 08, 2010 at 11:11:52AM +0100, Heiko Carstens wrote: >> On Sat, Mar 06, 2010 at 01:40:46PM +0530, Sachin Sant wrote: >>> With latest 33 git(2.6.33-git11 : 64096c1741...) on a POWER6 box >>> >>> type=2000 audit(1267853400.180:1): initialized >>> Kprobe smoke test started >>> ------------[ cut here ]------------ >>> Badness at kernel/kprobes.c:264 >>> NIP: c0000000006251e0 LR: c000000000625190 CTR: c00000000007914c >>> REGS: c0000000fecc3680 TRAP: 0700 Not tainted (2.6.33-git11-autotest) >>> MSR: 8000000000029032 <EE,ME,CE,IR,DR> CR: 24000048 XER: 2000000b >>> TASK = c0000000feca0000[1] 'swapper' THREAD: c0000000fecc0000 CPU: 2 >>> GPR00: 0000000000000001 c0000000fecc3900 c000000000b297b0 c0000000fc680000 >>> GPR04: 0000000000000004 0000000000000000 0000000024022024 c000000000a2b9d0 >>> GPR08: 0000000000004000 c0000000fc680004 0000000000010000 0000000000000004 >>> GPR12: 0000000022000024 c000000000bc2b00 0000000000051bc3 0000000000051aa1 >>> GPR16: 0000000000051bbb 0000000000d00000 c0000000008011f8 c0000000007f1ba1 >>> GPR20: 00000000015e87a8 c0000000008e87a8 c0000000fecc3cc8 c0000000fecc3cd0 >>> GPR24: c0000000fecc3cd8 c0000000fecc3cc0 c0000000fecc3be0 0000000000000000 >>> GPR28: 0000000000000000 c000000000a2b8b8 c000000000a94888 d000000000bd0004 >>> NIP [c0000000006251e0] .free_insn_slot+0x84/0x12c >>> LR [c000000000625190] .free_insn_slot+0x34/0x12c >>> Call Trace: >>> [c0000000fecc3900] [c000000000625190] .free_insn_slot+0x34/0x12c (unreliable) >>> [c0000000fecc3990] [c000000000622050] .arch_remove_kprobe+0x28/0x48 >>> [c0000000fecc3a10] [c000000000623f58] .__unregister_kprobe_bottom+0x28/0x8c >>> [c0000000fecc3aa0] [c00000000062419c] .unregister_kprobes+0xc0/0xf0 >>> [c0000000fecc3b40] [c0000000006241ec] .unregister_kprobe+0x20/0x30 >>> [c0000000fecc3bb0] [c0000000000e081c] .init_test_probes+0xc4/0x66c >>> [c0000000fecc3c50] [c0000000008c288c] .init_kprobes+0x1f0/0x230 >>> [c0000000fecc3e30] [c0000000000097a4] .do_one_initcall+0x88/0x1bc >>> [c0000000fecc3ee0] [c0000000008a0490] .kernel_init+0x220/0x2dc >>> [c0000000fecc3f90] [c00000000002c4d0] .kernel_thread+0x54/0x70 >>> Instruction dump: >>> 7c00f850 7c804b92 2fa40000 419c007c 7d0a5b92 7fa44000 409c0070 7d232214 >>> 88090020 68000002 7c0000d0 78000fe0 <0b000000> 2fbc0000 419e0044 8123001c >>> >>> 2.6.33-git10(64ba99267...) was OK. >>> >>> This WARN_ON was introduced by commit 4610ee1d36... >>> >>> kprobes: Introduce generic insn_slot framework >> >> FWIW, same on s390... > > This patch should fix it: > > When freeing the instruction slot, the arithmetic to calculate the index > of the slot in the page needs to account for the total size of the > instruction on the various architectures. > > Calculate the index correctly when freeing the out-of-line execution slot. > > Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Right, thank you Ananth! Acked-by: Masami Hiramatsu <mhiramat@redhat.com> > --- > Index: linux-8mar/kernel/kprobes.c > =================================================================== > --- linux-8mar.orig/kernel/kprobes.c 2010-03-08 17:10:33.000000000 +0530 > +++ linux-8mar/kernel/kprobes.c 2010-03-08 17:12:12.000000000 +0530 > @@ -259,7 +259,8 @@ > struct kprobe_insn_page *kip; > > list_for_each_entry(kip, &c->pages, list) { > - long idx = ((long)slot - (long)kip->insns) / c->insn_size; > + long idx = ((long)slot - (long)kip->insns) / > + (c->insn_size * sizeof(kprobe_opcode_t)); > if (idx >= 0 && idx < slots_per_page(c)) { > WARN_ON(kip->slot_used[idx] != SLOT_USED); > if (dirty) { > >
Index: linux-8mar/kernel/kprobes.c =================================================================== --- linux-8mar.orig/kernel/kprobes.c 2010-03-08 17:10:33.000000000 +0530 +++ linux-8mar/kernel/kprobes.c 2010-03-08 17:12:12.000000000 +0530 @@ -259,7 +259,8 @@ struct kprobe_insn_page *kip; list_for_each_entry(kip, &c->pages, list) { - long idx = ((long)slot - (long)kip->insns) / c->insn_size; + long idx = ((long)slot - (long)kip->insns) / + (c->insn_size * sizeof(kprobe_opcode_t)); if (idx >= 0 && idx < slots_per_page(c)) { WARN_ON(kip->slot_used[idx] != SLOT_USED); if (dirty) {