Re: [PATCH v2] powerpc/kprobes: Fix kprobe Oops happens in booke
On Mon, 9 Aug 2021 10:36:58 +0800, Pu Lehui wrote: > When using kprobe on powerpc booke series processor, Oops happens > as show bellow: > > / # echo "p:myprobe do_nanosleep" > /sys/kernel/debug/tracing/kprobe_events > / # echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable > / # sleep 1 > [ 50.076730] Oops: Exception in kernel mode, sig: 5 [#1] > [ 50.077017] BE PAGE_SIZE=4K SMP NR_CPUS=24 QEMU e500 > [ 50.077221] Modules linked in: > [ 50.077462] CPU: 0 PID: 77 Comm: sleep Not tainted > 5.14.0-rc4-00022-g251a1524293d #21 > [ 50.077887] NIP: c0b9c4e0 LR: c00ebecc CTR: > [ 50.078067] REGS: c3883de0 TRAP: 0700 Not tainted > (5.14.0-rc4-00022-g251a1524293d) > [ 50.078349] MSR: 00029000 CR: 24000228 XER: 2000 > [ 50.078675] > [ 50.078675] GPR00: c00ebdf0 c3883e90 c313e300 c3883ea0 0001 > c3883ecc 0001 > [ 50.078675] GPR08: c100598c c00ea250 0004 24000222 102490c2 > bff4180c 101e60d4 > [ 50.078675] GPR16: 102454ac 0040 1024 10241100 102410f8 > 1024 0050 > [ 50.078675] GPR24: 0002 c3883ea0 0001 c350 > 3b9b8d50 > [ 50.080151] NIP [c0b9c4e0] do_nanosleep+0x0/0x190 > [ 50.080352] LR [c00ebecc] hrtimer_nanosleep+0x14c/0x1e0 > [ 50.080638] Call Trace: > [ 50.080801] [c3883e90] [c00ebdf0] hrtimer_nanosleep+0x70/0x1e0 (unreliable) > [ 50.081110] [c3883f00] [c00ec004] sys_nanosleep_time32+0xa4/0x110 > [ 50.081336] [c3883f40] [c001509c] ret_from_syscall+0x0/0x28 > [ 50.081541] --- interrupt: c00 at 0x100a4d08 > [ 50.081749] NIP: 100a4d08 LR: 101b5234 CTR: 0003 > [ 50.081931] REGS: c3883f50 TRAP: 0c00 Not tainted > (5.14.0-rc4-00022-g251a1524293d) > [ 50.082183] MSR: 0002f902 CR: 24000222 XER: > [ 50.082457] > [ 50.082457] GPR00: 00a2 bf980040 1024b4d0 bf980084 bf980084 6400 > 00555345 fefefeff > [ 50.082457] GPR08: 7f7f7f7f 101e 0069 0003 28000422 102490c2 > bff4180c 101e60d4 > [ 50.082457] GPR16: 102454ac 0040 1024 10241100 102410f8 > 1024 0050 > [ 50.082457] GPR24: 0002 bf9803f4 1024 100039e0 > 102444e8 > [ 50.083789] NIP [100a4d08] 0x100a4d08 > [ 50.083917] LR [101b5234] 0x101b5234 > [ 50.084042] --- interrupt: c00 > [ 50.084238] Instruction dump: > [ 50.084483] 4bfffc40 6000 6000 6000 9421fff0 39400402 914200c0 > 38210010 > [ 50.084841] 4bfffc20 <7fe8> 7c0802a6 > 7c892378 93c10048 > [ 50.085487] ---[ end trace f6fffe98e2fa8f3e ]--- > [ 50.085678] > Trace/breakpoint trap > > [...] Applied to powerpc/fixes. [1/1] powerpc/kprobes: Fix kprobe Oops happens in booke https://git.kernel.org/powerpc/c/43e8f76006592cb1573a959aa287c45421066f9c cheers
Re: [PATCH v2] powerpc/kprobes: Fix kprobe Oops happens in booke
On 2021/8/11 13:31, Christophe Leroy wrote: Le 11/08/2021 à 04:53, Pu Lehui a écrit : Ping, serious problem here. All booke ppc will trigger Oops when perform kprobes related operations. As far as I can see it is in the fixes branch: https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/log/?h=fixes Thanks. On 2021/8/9 10:36, Pu Lehui wrote: When using kprobe on powerpc booke series processor, Oops happens as show bellow: / # echo "p:myprobe do_nanosleep" > /sys/kernel/debug/tracing/kprobe_events / # echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable / # sleep 1 [ 50.076730] Oops: Exception in kernel mode, sig: 5 [#1] [ 50.077017] BE PAGE_SIZE=4K SMP NR_CPUS=24 QEMU e500 [ 50.077221] Modules linked in: [ 50.077462] CPU: 0 PID: 77 Comm: sleep Not tainted 5.14.0-rc4-00022-g251a1524293d #21 [ 50.077887] NIP: c0b9c4e0 LR: c00ebecc CTR: [ 50.078067] REGS: c3883de0 TRAP: 0700 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.078349] MSR: 00029000 CR: 24000228 XER: 2000 [ 50.078675] [ 50.078675] GPR00: c00ebdf0 c3883e90 c313e300 c3883ea0 0001 c3883ecc 0001 [ 50.078675] GPR08: c100598c c00ea250 0004 24000222 102490c2 bff4180c 101e60d4 [ 50.078675] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.078675] GPR24: 0002 c3883ea0 0001 c350 3b9b8d50 [ 50.080151] NIP [c0b9c4e0] do_nanosleep+0x0/0x190 [ 50.080352] LR [c00ebecc] hrtimer_nanosleep+0x14c/0x1e0 [ 50.080638] Call Trace: [ 50.080801] [c3883e90] [c00ebdf0] hrtimer_nanosleep+0x70/0x1e0 (unreliable) [ 50.081110] [c3883f00] [c00ec004] sys_nanosleep_time32+0xa4/0x110 [ 50.081336] [c3883f40] [c001509c] ret_from_syscall+0x0/0x28 [ 50.081541] --- interrupt: c00 at 0x100a4d08 [ 50.081749] NIP: 100a4d08 LR: 101b5234 CTR: 0003 [ 50.081931] REGS: c3883f50 TRAP: 0c00 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.082183] MSR: 0002f902 CR: 24000222 XER: [ 50.082457] [ 50.082457] GPR00: 00a2 bf980040 1024b4d0 bf980084 bf980084 6400 00555345 fefefeff [ 50.082457] GPR08: 7f7f7f7f 101e 0069 0003 28000422 102490c2 bff4180c 101e60d4 [ 50.082457] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.082457] GPR24: 0002 bf9803f4 1024 100039e0 102444e8 [ 50.083789] NIP [100a4d08] 0x100a4d08 [ 50.083917] LR [101b5234] 0x101b5234 [ 50.084042] --- interrupt: c00 [ 50.084238] Instruction dump: [ 50.084483] 4bfffc40 6000 6000 6000 9421fff0 39400402 914200c0 38210010 [ 50.084841] 4bfffc20 <7fe8> 7c0802a6 7c892378 93c10048 [ 50.085487] ---[ end trace f6fffe98e2fa8f3e ]--- [ 50.085678] Trace/breakpoint trap There is no real mode for booke arch and the MMU translation is always on. The corresponding MSR_IS/MSR_DS bit in booke is used to switch the address space, but not for real mode judgment. Fixes: 21f8b2fa3ca5 ("powerpc/kprobes: Ignore traps that happened in real mode") Signed-off-by: Pu Lehui --- v1->v2: - use IS_ENABLED(CONFIG_BOOKE) as suggested by Michael Ellerman and Christophe Leroy - update Oops log to make problem clear arch/powerpc/kernel/kprobes.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/powerpc/kernel/kprobes.c b/arch/powerpc/kernel/kprobes.c index cbc28d1a2e1b..7a7cd6bda53e 100644 --- a/arch/powerpc/kernel/kprobes.c +++ b/arch/powerpc/kernel/kprobes.c @@ -292,7 +292,8 @@ int kprobe_handler(struct pt_regs *regs) if (user_mode(regs)) return 0; - if (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR)) + if (!IS_ENABLED(CONFIG_BOOKE) && + (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR))) return 0; /* .
Re: [PATCH v2] powerpc/kprobes: Fix kprobe Oops happens in booke
Le 11/08/2021 à 04:53, Pu Lehui a écrit : Ping, serious problem here. All booke ppc will trigger Oops when perform kprobes related operations. As far as I can see it is in the fixes branch: https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/log/?h=fixes On 2021/8/9 10:36, Pu Lehui wrote: When using kprobe on powerpc booke series processor, Oops happens as show bellow: / # echo "p:myprobe do_nanosleep" > /sys/kernel/debug/tracing/kprobe_events / # echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable / # sleep 1 [ 50.076730] Oops: Exception in kernel mode, sig: 5 [#1] [ 50.077017] BE PAGE_SIZE=4K SMP NR_CPUS=24 QEMU e500 [ 50.077221] Modules linked in: [ 50.077462] CPU: 0 PID: 77 Comm: sleep Not tainted 5.14.0-rc4-00022-g251a1524293d #21 [ 50.077887] NIP: c0b9c4e0 LR: c00ebecc CTR: [ 50.078067] REGS: c3883de0 TRAP: 0700 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.078349] MSR: 00029000 CR: 24000228 XER: 2000 [ 50.078675] [ 50.078675] GPR00: c00ebdf0 c3883e90 c313e300 c3883ea0 0001 c3883ecc 0001 [ 50.078675] GPR08: c100598c c00ea250 0004 24000222 102490c2 bff4180c 101e60d4 [ 50.078675] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.078675] GPR24: 0002 c3883ea0 0001 c350 3b9b8d50 [ 50.080151] NIP [c0b9c4e0] do_nanosleep+0x0/0x190 [ 50.080352] LR [c00ebecc] hrtimer_nanosleep+0x14c/0x1e0 [ 50.080638] Call Trace: [ 50.080801] [c3883e90] [c00ebdf0] hrtimer_nanosleep+0x70/0x1e0 (unreliable) [ 50.081110] [c3883f00] [c00ec004] sys_nanosleep_time32+0xa4/0x110 [ 50.081336] [c3883f40] [c001509c] ret_from_syscall+0x0/0x28 [ 50.081541] --- interrupt: c00 at 0x100a4d08 [ 50.081749] NIP: 100a4d08 LR: 101b5234 CTR: 0003 [ 50.081931] REGS: c3883f50 TRAP: 0c00 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.082183] MSR: 0002f902 CR: 24000222 XER: [ 50.082457] [ 50.082457] GPR00: 00a2 bf980040 1024b4d0 bf980084 bf980084 6400 00555345 fefefeff [ 50.082457] GPR08: 7f7f7f7f 101e 0069 0003 28000422 102490c2 bff4180c 101e60d4 [ 50.082457] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.082457] GPR24: 0002 bf9803f4 1024 100039e0 102444e8 [ 50.083789] NIP [100a4d08] 0x100a4d08 [ 50.083917] LR [101b5234] 0x101b5234 [ 50.084042] --- interrupt: c00 [ 50.084238] Instruction dump: [ 50.084483] 4bfffc40 6000 6000 6000 9421fff0 39400402 914200c0 38210010 [ 50.084841] 4bfffc20 <7fe8> 7c0802a6 7c892378 93c10048 [ 50.085487] ---[ end trace f6fffe98e2fa8f3e ]--- [ 50.085678] Trace/breakpoint trap There is no real mode for booke arch and the MMU translation is always on. The corresponding MSR_IS/MSR_DS bit in booke is used to switch the address space, but not for real mode judgment. Fixes: 21f8b2fa3ca5 ("powerpc/kprobes: Ignore traps that happened in real mode") Signed-off-by: Pu Lehui --- v1->v2: - use IS_ENABLED(CONFIG_BOOKE) as suggested by Michael Ellerman and Christophe Leroy - update Oops log to make problem clear arch/powerpc/kernel/kprobes.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/powerpc/kernel/kprobes.c b/arch/powerpc/kernel/kprobes.c index cbc28d1a2e1b..7a7cd6bda53e 100644 --- a/arch/powerpc/kernel/kprobes.c +++ b/arch/powerpc/kernel/kprobes.c @@ -292,7 +292,8 @@ int kprobe_handler(struct pt_regs *regs) if (user_mode(regs)) return 0; - if (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR)) + if (!IS_ENABLED(CONFIG_BOOKE) && + (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR))) return 0; /*
Re: [PATCH v2] powerpc/kprobes: Fix kprobe Oops happens in booke
Ping, serious problem here. All booke ppc will trigger Oops when perform kprobes related operations. On 2021/8/9 10:36, Pu Lehui wrote: When using kprobe on powerpc booke series processor, Oops happens as show bellow: / # echo "p:myprobe do_nanosleep" > /sys/kernel/debug/tracing/kprobe_events / # echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable / # sleep 1 [ 50.076730] Oops: Exception in kernel mode, sig: 5 [#1] [ 50.077017] BE PAGE_SIZE=4K SMP NR_CPUS=24 QEMU e500 [ 50.077221] Modules linked in: [ 50.077462] CPU: 0 PID: 77 Comm: sleep Not tainted 5.14.0-rc4-00022-g251a1524293d #21 [ 50.077887] NIP: c0b9c4e0 LR: c00ebecc CTR: [ 50.078067] REGS: c3883de0 TRAP: 0700 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.078349] MSR: 00029000 CR: 24000228 XER: 2000 [ 50.078675] [ 50.078675] GPR00: c00ebdf0 c3883e90 c313e300 c3883ea0 0001 c3883ecc 0001 [ 50.078675] GPR08: c100598c c00ea250 0004 24000222 102490c2 bff4180c 101e60d4 [ 50.078675] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.078675] GPR24: 0002 c3883ea0 0001 c350 3b9b8d50 [ 50.080151] NIP [c0b9c4e0] do_nanosleep+0x0/0x190 [ 50.080352] LR [c00ebecc] hrtimer_nanosleep+0x14c/0x1e0 [ 50.080638] Call Trace: [ 50.080801] [c3883e90] [c00ebdf0] hrtimer_nanosleep+0x70/0x1e0 (unreliable) [ 50.081110] [c3883f00] [c00ec004] sys_nanosleep_time32+0xa4/0x110 [ 50.081336] [c3883f40] [c001509c] ret_from_syscall+0x0/0x28 [ 50.081541] --- interrupt: c00 at 0x100a4d08 [ 50.081749] NIP: 100a4d08 LR: 101b5234 CTR: 0003 [ 50.081931] REGS: c3883f50 TRAP: 0c00 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.082183] MSR: 0002f902 CR: 24000222 XER: [ 50.082457] [ 50.082457] GPR00: 00a2 bf980040 1024b4d0 bf980084 bf980084 6400 00555345 fefefeff [ 50.082457] GPR08: 7f7f7f7f 101e 0069 0003 28000422 102490c2 bff4180c 101e60d4 [ 50.082457] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.082457] GPR24: 0002 bf9803f4 1024 100039e0 102444e8 [ 50.083789] NIP [100a4d08] 0x100a4d08 [ 50.083917] LR [101b5234] 0x101b5234 [ 50.084042] --- interrupt: c00 [ 50.084238] Instruction dump: [ 50.084483] 4bfffc40 6000 6000 6000 9421fff0 39400402 914200c0 38210010 [ 50.084841] 4bfffc20 <7fe8> 7c0802a6 7c892378 93c10048 [ 50.085487] ---[ end trace f6fffe98e2fa8f3e ]--- [ 50.085678] Trace/breakpoint trap There is no real mode for booke arch and the MMU translation is always on. The corresponding MSR_IS/MSR_DS bit in booke is used to switch the address space, but not for real mode judgment. Fixes: 21f8b2fa3ca5 ("powerpc/kprobes: Ignore traps that happened in real mode") Signed-off-by: Pu Lehui --- v1->v2: - use IS_ENABLED(CONFIG_BOOKE) as suggested by Michael Ellerman and Christophe Leroy - update Oops log to make problem clear arch/powerpc/kernel/kprobes.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/powerpc/kernel/kprobes.c b/arch/powerpc/kernel/kprobes.c index cbc28d1a2e1b..7a7cd6bda53e 100644 --- a/arch/powerpc/kernel/kprobes.c +++ b/arch/powerpc/kernel/kprobes.c @@ -292,7 +292,8 @@ int kprobe_handler(struct pt_regs *regs) if (user_mode(regs)) return 0; - if (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR)) + if (!IS_ENABLED(CONFIG_BOOKE) && + (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR))) return 0; /*
[PATCH v2] powerpc/kprobes: Fix kprobe Oops happens in booke
When using kprobe on powerpc booke series processor, Oops happens as show bellow: / # echo "p:myprobe do_nanosleep" > /sys/kernel/debug/tracing/kprobe_events / # echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable / # sleep 1 [ 50.076730] Oops: Exception in kernel mode, sig: 5 [#1] [ 50.077017] BE PAGE_SIZE=4K SMP NR_CPUS=24 QEMU e500 [ 50.077221] Modules linked in: [ 50.077462] CPU: 0 PID: 77 Comm: sleep Not tainted 5.14.0-rc4-00022-g251a1524293d #21 [ 50.077887] NIP: c0b9c4e0 LR: c00ebecc CTR: [ 50.078067] REGS: c3883de0 TRAP: 0700 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.078349] MSR: 00029000 CR: 24000228 XER: 2000 [ 50.078675] [ 50.078675] GPR00: c00ebdf0 c3883e90 c313e300 c3883ea0 0001 c3883ecc 0001 [ 50.078675] GPR08: c100598c c00ea250 0004 24000222 102490c2 bff4180c 101e60d4 [ 50.078675] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.078675] GPR24: 0002 c3883ea0 0001 c350 3b9b8d50 [ 50.080151] NIP [c0b9c4e0] do_nanosleep+0x0/0x190 [ 50.080352] LR [c00ebecc] hrtimer_nanosleep+0x14c/0x1e0 [ 50.080638] Call Trace: [ 50.080801] [c3883e90] [c00ebdf0] hrtimer_nanosleep+0x70/0x1e0 (unreliable) [ 50.081110] [c3883f00] [c00ec004] sys_nanosleep_time32+0xa4/0x110 [ 50.081336] [c3883f40] [c001509c] ret_from_syscall+0x0/0x28 [ 50.081541] --- interrupt: c00 at 0x100a4d08 [ 50.081749] NIP: 100a4d08 LR: 101b5234 CTR: 0003 [ 50.081931] REGS: c3883f50 TRAP: 0c00 Not tainted (5.14.0-rc4-00022-g251a1524293d) [ 50.082183] MSR: 0002f902 CR: 24000222 XER: [ 50.082457] [ 50.082457] GPR00: 00a2 bf980040 1024b4d0 bf980084 bf980084 6400 00555345 fefefeff [ 50.082457] GPR08: 7f7f7f7f 101e 0069 0003 28000422 102490c2 bff4180c 101e60d4 [ 50.082457] GPR16: 102454ac 0040 1024 10241100 102410f8 1024 0050 [ 50.082457] GPR24: 0002 bf9803f4 1024 100039e0 102444e8 [ 50.083789] NIP [100a4d08] 0x100a4d08 [ 50.083917] LR [101b5234] 0x101b5234 [ 50.084042] --- interrupt: c00 [ 50.084238] Instruction dump: [ 50.084483] 4bfffc40 6000 6000 6000 9421fff0 39400402 914200c0 38210010 [ 50.084841] 4bfffc20 <7fe8> 7c0802a6 7c892378 93c10048 [ 50.085487] ---[ end trace f6fffe98e2fa8f3e ]--- [ 50.085678] Trace/breakpoint trap There is no real mode for booke arch and the MMU translation is always on. The corresponding MSR_IS/MSR_DS bit in booke is used to switch the address space, but not for real mode judgment. Fixes: 21f8b2fa3ca5 ("powerpc/kprobes: Ignore traps that happened in real mode") Signed-off-by: Pu Lehui --- v1->v2: - use IS_ENABLED(CONFIG_BOOKE) as suggested by Michael Ellerman and Christophe Leroy - update Oops log to make problem clear arch/powerpc/kernel/kprobes.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/powerpc/kernel/kprobes.c b/arch/powerpc/kernel/kprobes.c index cbc28d1a2e1b..7a7cd6bda53e 100644 --- a/arch/powerpc/kernel/kprobes.c +++ b/arch/powerpc/kernel/kprobes.c @@ -292,7 +292,8 @@ int kprobe_handler(struct pt_regs *regs) if (user_mode(regs)) return 0; - if (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR)) + if (!IS_ENABLED(CONFIG_BOOKE) && + (!(regs->msr & MSR_IR) || !(regs->msr & MSR_DR))) return 0; /* -- 2.17.1