| Message ID | 20210227091701.23944-1-huangshaobo6@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [4.4.y] arm: kprobes: Allow to handle reentered kprobe on single-stepping | expand |
> > From: Masami Hiramatsu <mhiramat@kernel.org> > > commit f3fbd7ec62dec1528fb8044034e2885f2b257941 upstream > > This is arm port of commit 6a5022a56ac3 ("kprobes/x86: Allow to > handle reentered kprobe on single-stepping") > > Since the FIQ handlers can interrupt in the single stepping > (or preparing the single stepping, do_debug etc.), we should > consider a kprobe is hit in the NMI handler. Even in that > case, the kprobe is allowed to be reentered as same as the > kprobes hit in kprobe handlers > (KPROBE_HIT_ACTIVE or KPROBE_HIT_SSDONE). > > The real issue will happen when a kprobe hit while another > reentered kprobe is processing (KPROBE_REENTER), because > we already consumed a saved-area for the previous kprobe. > > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> > Signed-off-by: Jon Medhurst <tixy@linaro.org> > Fixes: 24ba613c9d6c ("ARM kprobes: core code") > Cc: stable@vger.kernel.org #v2.6.25~v4.11 > Signed-off-by: huangshaobo <huangshaobo6@huawei.com> arm32 jprobe does not handle interrupt reentry correctly. When jprobe is running in singlestep, jprobe is triggered again in system interrupt, it will run to the BUG function to cause panic. jprobe needs to enter the kprobe_handler function twice, the first time to enter kprobe_handler to save the context environment; the second time to enter kprobe_handler to simulate the replaced instruction and restore the context environment of the probed function.But the system interrupt is not closed when the kprobe_handler is entered for the second time in jprobe_return. the code of jprobe_return is as follows: void __kprobes jprobe_return(void) { struct kprobe_ctlblk *kcb = get_kprobe_ctlblk(); __asm__ __volatile__ ( /* * Setup an empty pt_regs. Fill SP and PC fields as * they're needed by longjmp_break_handler. * * We allocate some slack between the original SP and start of * our fabricated regs. To be precise we want to have worst case * covered which is STMFD with all 16 regs so we allocate 2 * * sizeof(struct_pt_regs)). * * This is to prevent any simulated instruction from writing * over the regs when they are accessing the stack. */ #ifdef CONFIG_THUMB2_KERNEL ... #else "sub sp, %0, %1 \n\t" #endif "ldr r0, ="__stringify(JPROBE_MAGIC_ADDR)"\n\t" "str %0, [sp, %2] \n\t" "str r0, [sp, %3] \n\t" "mov r0, sp \n\t" "bl kprobe_handler \n\t" /* * Return to the context saved by setjmp_pre_handler * and restored by longjmp_break_handler. */ #ifdef CONFIG_THUMB2_KERNEL ... #else "ldr r0, [sp, %4] \n\t" "msr cpsr_cxsf, r0 \n\t" "ldmia sp, {r0 - pc} \n\t" #endif : : "r" (kcb->jprobe_saved_regs.ARM_sp), "I" (sizeof(struct pt_regs) * 2), "J" (offsetof(struct pt_regs, ARM_sp)), "J" (offsetof(struct pt_regs, ARM_pc)), "J" (offsetof(struct pt_regs, ARM_cpsr)), "J" (offsetof(struct pt_regs, ARM_lr)) : "memory", "cc"); } If the execution reaches the singlestep, the jprobe status is KPROBE_HIT_SS. If an interrupt is generated at this time, and jprobe is triggered again in the interrupt, jprobe reentry will be generated, at this time kcb->kprobe_status==KPROBE_HIT_SS, Enter the default branch to execute the BUG function, causing the system to panic. The partial code of kprobe_handler is as follows: void __kprobes kprobe_handler(struct pt_regs *regs) { struct kprobe *p, *cur; struct kprobe_ctlblk *kcb; kcb = get_kprobe_ctlblk(); cur = kprobe_running(); #ifdef CONFIG_THUMB2_KERNEL /* * First look for a probe which was registered using an address with * bit 0 set, this is the usual situation for pointers to Thumb code. * If not found, fallback to looking for one with bit 0 clear. */ p = get_kprobe((kprobe_opcode_t *)(regs->ARM_pc | 1)); if (!p) p = get_kprobe((kprobe_opcode_t *)regs->ARM_pc); #else /* ! CONFIG_THUMB2_KERNEL */ p = get_kprobe((kprobe_opcode_t *)regs->ARM_pc); #endif if (p) { if (cur) { /* Kprobe is pending, so we're recursing. */ switch (kcb->kprobe_status) { case KPROBE_HIT_ACTIVE: case KPROBE_HIT_SSDONE: /* A pre- or post-handler probe got us here. */ kprobes_inc_nmissed_count(p); save_previous_kprobe(kcb); set_current_kprobe(p); kcb->kprobe_status = KPROBE_REENTER; singlestep(p, regs, kcb); restore_previous_kprobe(kcb); break; default: /* impossible cases */ BUG(); …… } At this time, it is necessary to increase the processing of the KPROBE_HIT_SS state, that is, jprobe only executes the simulation instruction, and does not execute the user callback function. For example, use jprobe to register arm_dma_map_page, arm_dma_map_page will be called by gmac_recv in the interrupt, the system will panic, and the log is as follows: kernel BUG at arch/arm/probes/kprobes/core.c:296! Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM Modules linked in:…… CPU: 0 PID: 0 Comm: swapper/0 Tainted: G O 4.4.236 #15 Hardware name: Hisilicon A15 task: c0a25220 task.stack: c0a00000 PC is at kprobe_handler+0x120/0x288 LR is at kprobe_handler+0x120/0x288 pc : [<c0597d24>] lr : [<c0597d24>] psr: 600f0193 sp : c0a01b08 ip : 00000007 fp : c0602088 r10: eeb1a26c r9 : 2e2b1000 r8 : c0a32530 r7 : c0a01be8 r6 : c086926c r5 : bf4c3000 r4 : c0869268 r3 : c0a2244c r2 : 00000000 r1 : 00000104 r0 : 00000032 Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment kernel Control: 30c5387d Table: 2c1dfd00 DAC: 55555555 Process swapper/0 (pid: 0, stack limit = 0xc0a00210) Stack: (0xc0a01b08 to 0xc0a02000) …… [<c0597d24>] (kprobe_handler) from [<c0597eb0>] (kprobe_trap_handler+0x24/0x48) [<c0597eb0>] (kprobe_trap_handler) from [<c02011ac>] (do_undefinstr+0x1a8/0x234) [<c02011ac>] (do_undefinstr) from [<c0596b30>] (__und_svc_finish+0x0/0x30) Exception stack(0xc0a01be8 to 0xc0a01c30) 1be0: ed8523f8 ef1cdc20 00000540 00000080 eaac8540 00000080 1c00: c02159ec c0a265e0 00000002 00000000 bf3f9000 c0602088 0002b4c8 c0a01c78 1c20: bf3e4b38 c02159ec a00f0113 ffffffff [<c0596b30>] (__und_svc_finish) from [<c02159ec>] (arm_dma_map_page+0x0/0xe4) [<c02159ec>] (arm_dma_map_page) from [<c0597e54>] (kprobe_handler+0x250/0x288) [<c0597e54>] (kprobe_handler) from [<bf3e56cc>] (gmac_recv+0xc0/0x858 [Drv_Gmac_K]) [<bf3e56cc>] (gmac_recv [Drv_Gmac_K]) from [<c04f96e4>] (net_rx_action+0x130/0x3d4) [<c04f96e4>] (net_rx_action) from [<c02245bc>] (__do_softirq+0x1b0/0x418) [<c02245bc>] (__do_softirq) from [<c0224b4c>] (irq_exit+0xc0/0x188) [<c0224b4c>] (irq_exit) from [<c0247604>] (msa_irq_exit+0x130/0x220) [<c0247604>] (msa_irq_exit) from [<c027283c>] (__handle_domain_irq+0xec/0x1b4) [<c027283c>] (__handle_domain_irq) from [<c0201580>] (gic_handle_irq+0xcc/0x104) [<c0201580>] (gic_handle_irq) from [<c0596a40>] (__irq_svc+0x40/0x58) Exception stack(0xc0a01e78 to 0xc0a01ec0) 1e60: 00000000 c0a32530 1e80: 2e2b1000 c0869268 c0a72500 00000000 e9a7f4a8 c0a25220 c0591b08 00000040 1ea0: c0a22a34 c0a01f34 00000000 c0a01ec8 c02c08e0 c044a498 200f0013 ffffffff [<c0596a40>] (__irq_svc) from [<c044a498>] (debug_smp_processor_id+0x0/0x14) [<c044a498>] (debug_smp_processor_id) from [<c02c08e0>] (kprobe_busy_begin+0x38/0x6c) [<c02c08e0>] (kprobe_busy_begin) from [<c02c09bc>] (kprobe_flush_task+0x38/0x120) [<c02c09bc>] (kprobe_flush_task) from [<c024be9c>] (finish_task_switch+0x19c/0x2a4) [<c024be9c>] (finish_task_switch) from [<c0591b08>] (__schedule+0x5a4/0x6b4) [<c0591b08>] (__schedule) from [<c0591e00>] (schedule+0x98/0xc4) [<c0591e00>] (schedule) from [<c05920b0>] (schedule_preempt_disabled+0x14/0x20) [<c05920b0>] (schedule_preempt_disabled) from [<c026692c>] (cpu_startup_entry+0x3c/0x33c) [<c026692c>] (cpu_startup_entry) from [<c058f030>] (rest_init+0xac/0xf0) [<c058f030>] (rest_init) from [<c0800cb8>] (start_kernel+0x484/0x534) Code: e59f016c ebf57071 e1a00005 ebf49e8a (e7f001f2)
On Sat, Feb 27, 2021 at 05:17:01PM +0800, huangshaobo wrote: > From: Masami Hiramatsu <mhiramat@kernel.org> > > commit f3fbd7ec62dec1528fb8044034e2885f2b257941 upstream > > This is arm port of commit 6a5022a56ac3 ("kprobes/x86: Allow to > handle reentered kprobe on single-stepping") > > Since the FIQ handlers can interrupt in the single stepping > (or preparing the single stepping, do_debug etc.), we should > consider a kprobe is hit in the NMI handler. Even in that > case, the kprobe is allowed to be reentered as same as the > kprobes hit in kprobe handlers > (KPROBE_HIT_ACTIVE or KPROBE_HIT_SSDONE). > > The real issue will happen when a kprobe hit while another > reentered kprobe is processing (KPROBE_REENTER), because > we already consumed a saved-area for the previous kprobe. > > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> > Signed-off-by: Jon Medhurst <tixy@linaro.org> > Fixes: 24ba613c9d6c ("ARM kprobes: core code") > Cc: stable@vger.kernel.org #v2.6.25~v4.11 > Signed-off-by: huangshaobo <huangshaobo6@huawei.com> > --- > arch/arm/probes/kprobes/core.c | 6 ++++++ > 1 file changed, 6 insertions(+) What about the 4.9.y tree as well? thanks, greg k-h
On March 1, 2021 at 11:30 AM, Greg KH wrote: > On Mon, Feb 27, 2021 at 05:17:01PM +0800, huangshaobo wrote: > > From: Masami Hiramatsu <mhiramat@kernel.org> > > > > commit f3fbd7ec62dec1528fb8044034e2885f2b257941 upstream > > > > This is arm port of commit 6a5022a56ac3 ("kprobes/x86: Allow to handle > > reentered kprobe on single-stepping") > > > > Since the FIQ handlers can interrupt in the single stepping (or > > preparing the single stepping, do_debug etc.), we should consider a > > kprobe is hit in the NMI handler. Even in that case, the kprobe is > > allowed to be reentered as same as the kprobes hit in kprobe handlers > > (KPROBE_HIT_ACTIVE or KPROBE_HIT_SSDONE). > > > > The real issue will happen when a kprobe hit while another reentered > > kprobe is processing (KPROBE_REENTER), because we already consumed a > > saved-area for the previous kprobe. > > > > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> > > Signed-off-by: Jon Medhurst <tixy@linaro.org> > > Fixes: 24ba613c9d6c ("ARM kprobes: core code") > > Cc: stable@vger.kernel.org #v2.6.25~v4.11 > > Signed-off-by: huangshaobo <huangshaobo6@huawei.com> > > --- > > arch/arm/probes/kprobes/core.c | 6 ++++++ > > 1 file changed, 6 insertions(+) > > What about the 4.9.y tree as well? > > thanks, > > greg k-h Yes, I tested on the 4.4.y tree. From the code analysis, the same problem exists in the 2.6.25 to 4.11 trees, and of course the 4.9.y tree is also included. thanks, ShaoBo Huang
ping On Sat, Feb 27, 2021 at 05:17:01PM +0800, huangshaobo wrote: > From: Masami Hiramatsu <mhiramat@kernel.org> > > commit f3fbd7ec62dec1528fb8044034e2885f2b257941 upstream > > This is arm port of commit 6a5022a56ac3 ("kprobes/x86: Allow to > handle reentered kprobe on single-stepping") > > Since the FIQ handlers can interrupt in the single stepping > (or preparing the single stepping, do_debug etc.), we should > consider a kprobe is hit in the NMI handler. Even in that > case, the kprobe is allowed to be reentered as same as the > kprobes hit in kprobe handlers > (KPROBE_HIT_ACTIVE or KPROBE_HIT_SSDONE). > > The real issue will happen when a kprobe hit while another > reentered kprobe is processing (KPROBE_REENTER), because > we already consumed a saved-area for the previous kprobe. > > Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org> > Signed-off-by: Jon Medhurst <tixy@linaro.org> > Fixes: 24ba613c9d6c ("ARM kprobes: core code") > Cc: stable@vger.kernel.org #v2.6.25~v4.11 > Signed-off-by: huangshaobo <huangshaobo6@huawei.com> > --- > arch/arm/probes/kprobes/core.c | 6 ++++++ > 1 file changed, 6 insertions(+)
On Thu, Jul 08, 2021 at 08:42:40PM +0800, Shaobo Huang wrote:
> ping
???
What are you asking about here? Please be specific, "naked pings" like
this are not informative.
thanks,
greg k-h
On Thu, Jul 08, 2021 at 09:56PM +0800, Greg KH wrote: > > On Thu, Jul 08, 2021 at 08:42:40PM +0800, Shaobo Huang wrote: > > ping > > ??? > > What are you asking about here? Please be specific, "naked pings" like this are not informative. > > thanks, > > greg k-h This patch was merged in the 4.9.y branch last time, but the 4.4.y branch is still not merged, please review the 4.4 branch merge. thanks, ShaoBo Huang
On Thu, Jul 08, 2021 at 09:12:40PM +0800, Shaobo Huang wrote: > On Thu, Jul 08, 2021 at 09:56PM +0800, Greg KH wrote: > > > On Thu, Jul 08, 2021 at 08:42:40PM +0800, Shaobo Huang wrote: > > > ping > > > > ??? > > > > What are you asking about here? Please be specific, "naked pings" like this are not informative. > > > > thanks, > > > > greg k-h > > This patch was merged in the 4.9.y branch last time, but the 4.4.y branch is still not merged, > please review the 4.4 branch merge. What "4.4 branch merge"? I see no patch here at all. confused, greg k-h
diff --git a/arch/arm/probes/kprobes/core.c b/arch/arm/probes/kprobes/core.c index 3eb018fa1a1f..c3362ddd6c4c 100644 --- a/arch/arm/probes/kprobes/core.c +++ b/arch/arm/probes/kprobes/core.c @@ -270,6 +270,7 @@ void __kprobes kprobe_handler(struct pt_regs *regs) switch (kcb->kprobe_status) { case KPROBE_HIT_ACTIVE: case KPROBE_HIT_SSDONE: + case KPROBE_HIT_SS: /* A pre- or post-handler probe got us here. */ kprobes_inc_nmissed_count(p); save_previous_kprobe(kcb); @@ -278,6 +279,11 @@ void __kprobes kprobe_handler(struct pt_regs *regs) singlestep(p, regs, kcb); restore_previous_kprobe(kcb); break; + case KPROBE_REENTER: + /* A nested probe was hit in FIQ, it is a BUG */ + pr_warn("Unrecoverable kprobe detected at %p.\n", + p->addr); + /* fall through */ default: /* impossible cases */ BUG();