| Message ID | 20220427224924.592546-2-gpiccoli@igalia.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | The panic notifiers refactor | expand |
| Context | Check | Description |
|---|---|---|
| netdev/tree_selection | success | Not a local patch, async |
On 27/04/2022 19:48, Guilherme G. Piccoli wrote: > In the panic path we have a list of functions to be called, the panic > notifiers - such callbacks perform various actions in the machine's > last breath, and sometimes users want them to run before kdump. We > have the parameter "crash_kexec_post_notifiers" for that. When such > parameter is used, the function "crash_smp_send_stop()" is executed > to poweroff all secondary CPUs through the NMI-shootdown mechanism; > part of this process involves disabling virtualization features in > all CPUs (except the main one). > > Now, in the emergency restart procedure we have also a way of > disabling VMX in all CPUs, using the same NMI-shootdown mechanism; > what happens though is that in case we already NMI-disabled all CPUs, > the emergency restart fails due to a second addition of the same items > in the NMI list, as per the following log output: > > sysrq: Trigger a crash > Kernel panic - not syncing: sysrq triggered crash > [...] > Rebooting in 2 seconds.. > list_add double add: new=<addr1>, prev=<addr2>, next=<addr1>. > ------------[ cut here ]------------ > kernel BUG at lib/list_debug.c:29! > invalid opcode: 0000 [#1] PREEMPT SMP PTI > > In order to reproduce the problem, users just need to set the kernel > parameter "crash_kexec_post_notifiers" *without* kdump set in any > system with the VMX feature present. > > Since there is no benefit in re-disabling VMX in all CPUs in case > it was already done, this patch prevents that by guarding the restart > routine against doubly issuing NMIs unnecessarily. Notice we still > need to disable VMX locally in the emergency restart. > > Fixes: ed72736183c4 ("x86/reboot: Force all cpus to exit VMX root if VMX is supported) > Fixes: 0ee59413c967 ("x86/panic: replace smp_send_stop() with kdump friendly version in panic path") > Cc: David P. Reed <dpreed@deepplum.com> > Cc: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> > Cc: Paolo Bonzini <pbonzini@redhat.com> > Cc: Sean Christopherson <seanjc@google.com> > Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com> > --- > arch/x86/include/asm/cpu.h | 1 + > arch/x86/kernel/crash.c | 8 ++++---- > arch/x86/kernel/reboot.c | 14 ++++++++++++-- > 3 files changed, 17 insertions(+), 6 deletions(-) > Hi Paolo / Sean / Vitaly, sorry for the ping. But do you think this fix is OK from the VMX point-of-view? I'd like to send a V2 of this set soon, so any review here is highly appreciated! Cheers, Guilherme
I find the shortlog to be very confusing, the bug has nothing to do with disabling VMX and I distinctly remember wrapping VMXOFF with exception fixup to prevent doom if VMX is already disabled :-). The issue is really that nmi_shootdown_cpus() doesn't play nice with being called twice. On Wed, Apr 27, 2022, Guilherme G. Piccoli wrote: > In the panic path we have a list of functions to be called, the panic > notifiers - such callbacks perform various actions in the machine's > last breath, and sometimes users want them to run before kdump. We > have the parameter "crash_kexec_post_notifiers" for that. When such > parameter is used, the function "crash_smp_send_stop()" is executed > to poweroff all secondary CPUs through the NMI-shootdown mechanism; > part of this process involves disabling virtualization features in > all CPUs (except the main one). > > Now, in the emergency restart procedure we have also a way of > disabling VMX in all CPUs, using the same NMI-shootdown mechanism; > what happens though is that in case we already NMI-disabled all CPUs, > the emergency restart fails due to a second addition of the same items > in the NMI list, as per the following log output: > > sysrq: Trigger a crash > Kernel panic - not syncing: sysrq triggered crash > [...] > Rebooting in 2 seconds.. > list_add double add: new=<addr1>, prev=<addr2>, next=<addr1>. > ------------[ cut here ]------------ > kernel BUG at lib/list_debug.c:29! > invalid opcode: 0000 [#1] PREEMPT SMP PTI Call stacks for the two callers would be very, very helpful. > In order to reproduce the problem, users just need to set the kernel > parameter "crash_kexec_post_notifiers" *without* kdump set in any > system with the VMX feature present. > > Since there is no benefit in re-disabling VMX in all CPUs in case > it was already done, this patch prevents that by guarding the restart > routine against doubly issuing NMIs unnecessarily. Notice we still > need to disable VMX locally in the emergency restart. > > Fixes: ed72736183c4 ("x86/reboot: Force all cpus to exit VMX root if VMX is supported) > Fixes: 0ee59413c967 ("x86/panic: replace smp_send_stop() with kdump friendly version in panic path") > Cc: David P. Reed <dpreed@deepplum.com> > Cc: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> > Cc: Paolo Bonzini <pbonzini@redhat.com> > Cc: Sean Christopherson <seanjc@google.com> > Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com> > --- > arch/x86/include/asm/cpu.h | 1 + > arch/x86/kernel/crash.c | 8 ++++---- > arch/x86/kernel/reboot.c | 14 ++++++++++++-- > 3 files changed, 17 insertions(+), 6 deletions(-) > > diff --git a/arch/x86/include/asm/cpu.h b/arch/x86/include/asm/cpu.h > index 86e5e4e26fcb..b6a9062d387f 100644 > --- a/arch/x86/include/asm/cpu.h > +++ b/arch/x86/include/asm/cpu.h > @@ -36,6 +36,7 @@ extern int _debug_hotplug_cpu(int cpu, int action); > #endif > #endif > > +extern bool crash_cpus_stopped; > int mwait_usable(const struct cpuinfo_x86 *); > > unsigned int x86_family(unsigned int sig); > diff --git a/arch/x86/kernel/crash.c b/arch/x86/kernel/crash.c > index e8326a8d1c5d..71dd1a990e8d 100644 > --- a/arch/x86/kernel/crash.c > +++ b/arch/x86/kernel/crash.c > @@ -42,6 +42,8 @@ > #include <asm/crash.h> > #include <asm/cmdline.h> > > +bool crash_cpus_stopped; > + > /* Used while preparing memory map entries for second kernel */ > struct crash_memmap_data { > struct boot_params *params; > @@ -108,9 +110,7 @@ void kdump_nmi_shootdown_cpus(void) > /* Override the weak function in kernel/panic.c */ > void crash_smp_send_stop(void) > { > - static int cpus_stopped; > - > - if (cpus_stopped) > + if (crash_cpus_stopped) > return; > > if (smp_ops.crash_stop_other_cpus) > @@ -118,7 +118,7 @@ void crash_smp_send_stop(void) > else > smp_send_stop(); > > - cpus_stopped = 1; > + crash_cpus_stopped = true; This feels like were just adding more duct tape to the mess. nmi_shootdown() is still unsafe for more than one caller, and it takes a _lot_ of staring and searching to understand that crash_smp_send_stop() is invoked iff CONFIG_KEXEC_CORE=y, i.e. that it will call smp_ops.crash_stop_other_cpus() and not just smp_send_stop(). Rather than shared a flag between two relatively unrelated functions, what if we instead disabling virtualization in crash_nmi_callback() and then turn the reboot call into a nop if an NMI shootdown has already occurred? That will also add a bit of documentation about multiple shootdowns not working. And I believe there's also a lurking bug in native_machine_emergency_restart() that can be fixed with cleanup. SVM can also block INIT and so should be disabled during an emergency reboot. The attached patches are compile tested only. If they seem sane, I'll post an official mini series. > } > > #else > diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c > index fa700b46588e..2fc42b8402ac 100644 > --- a/arch/x86/kernel/reboot.c > +++ b/arch/x86/kernel/reboot.c > @@ -589,8 +589,18 @@ static void native_machine_emergency_restart(void) > int orig_reboot_type = reboot_type; > unsigned short mode; > > - if (reboot_emergency) > - emergency_vmx_disable_all(); > + /* > + * We can reach this point in the end of panic path, having > + * NMI-disabled all secondary CPUs. This process involves > + * disabling the CPU virtualization technologies, so if that > + * is the case, we only miss disabling the local CPU VMX... > + */ > + if (reboot_emergency) { > + if (!crash_cpus_stopped) > + emergency_vmx_disable_all(); > + else > + cpu_emergency_vmxoff(); > + } > > tboot_shutdown(TB_SHUTDOWN_REBOOT); > > -- > 2.36.0 >
On 09/05/2022 12:52, Sean Christopherson wrote: > I find the shortlog to be very confusing, the bug has nothing to do with disabling > VMX and I distinctly remember wrapping VMXOFF with exception fixup to prevent doom > if VMX is already disabled :-). The issue is really that nmi_shootdown_cpus() doesn't > play nice with being called twice. > Hey Sean, OK - I agree with you, the issue is really about the double list addition. > [...] > > Call stacks for the two callers would be very, very helpful. > [...] > This feels like were just adding more duct tape to the mess. nmi_shootdown() is > still unsafe for more than one caller, and it takes a _lot_ of staring and searching > to understand that crash_smp_send_stop() is invoked iff CONFIG_KEXEC_CORE=y, i.e. > that it will call smp_ops.crash_stop_other_cpus() and not just smp_send_stop(). > > Rather than shared a flag between two relatively unrelated functions, what if we > instead disabling virtualization in crash_nmi_callback() and then turn the reboot > call into a nop if an NMI shootdown has already occurred? That will also add a > bit of documentation about multiple shootdowns not working. > > And I believe there's also a lurking bug in native_machine_emergency_restart() that > can be fixed with cleanup. SVM can also block INIT and so should be disabled during > an emergency reboot. > > The attached patches are compile tested only. If they seem sane, I'll post an > official mini series. Thanks Sean, it makes sense - my patch is more a "band-aid" whereas yours fixes it in a more generic way. Confess I found the logic of your patch complex, but as you said, it requires a *lot* of code analysis to understand these multiple shutdown patches, the problem is complicated by nature heh I've tested your patch 0001 and it works well for all cases [0], so go ahead and submit the miniseries, feel free to add: Reported-and-tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com> I've read patch 0002 and it makes sense to me as well, a good proactive bug fix =) With that said, I'll of course drop this one from V2 of this series. Cheers, Guilherme [0] A summary of my tests and the code paths that the panic shutdown take depending on some conditions: New function that disables VMX/SVM: cpu_crash_disable_virtualization() [should be executed in every online CPU on shutdown) The panic path triggers the following call stacks depending on kdump and post_notifiers: (1) kexec/kdump + !crash_kexec_post_notifiers ->machine_crash_shutdown() ----.crash_shutdown() <custom handler> ------native_machine_crash_shutdown() [all custom handlers except Xen PV call the native generic function] --------crash_smp_send_stop() ----------kdump_nmi_shootdown_cpus() ------------nmi_shootdown_cpus(kdump_nmi_callback) --------------crash_nmi_callback() ----------------kdump_nmi_callback() ------------------cpu_crash_disable_virtualization() (2) kexec/kdump + crash_kexec_post_notifiers ->crash_smp_send_stop() ----kdump_nmi_shootdown_cpus() ------nmi_shootdown_cpus(kdump_nmi_callback) --------crash_nmi_callback() ----------kdump_nmi_callback() ------------cpu_crash_disable_virtualization() After this path, will execute machine_crash_shutdown() but crash_smp_send_stop() is guarded against double execution. Also, emergency restart calls emergency_vmx_disable_all() . (3) !kexec/kdump + crash_kexec_post_notifiers Same as (2) (4) !kexec/kdump + !crash_kexec_post_notifiers -> smp_send_stop() ----native_stop_other_cpus() ------apic_send_IPI_allbutself(REBOOT_VECTOR) --------sysvec_reboot ----------cpu_emergency_vmxoff() <if the IPI approach succeeded, CPU stopped here> If not: ------register_stop_handler() --------apic_send_IPI_allbutself(NMI_VECTOR) ----------smp_stop_nmi_callback() ------------cpu_emergency_vmxoff() After that, emergency_vmx_disable_all() gets called in the emergency restart path as well.
diff --git a/arch/x86/include/asm/cpu.h b/arch/x86/include/asm/cpu.h index 86e5e4e26fcb..b6a9062d387f 100644 --- a/arch/x86/include/asm/cpu.h +++ b/arch/x86/include/asm/cpu.h @@ -36,6 +36,7 @@ extern int _debug_hotplug_cpu(int cpu, int action); #endif #endif +extern bool crash_cpus_stopped; int mwait_usable(const struct cpuinfo_x86 *); unsigned int x86_family(unsigned int sig); diff --git a/arch/x86/kernel/crash.c b/arch/x86/kernel/crash.c index e8326a8d1c5d..71dd1a990e8d 100644 --- a/arch/x86/kernel/crash.c +++ b/arch/x86/kernel/crash.c @@ -42,6 +42,8 @@ #include <asm/crash.h> #include <asm/cmdline.h> +bool crash_cpus_stopped; + /* Used while preparing memory map entries for second kernel */ struct crash_memmap_data { struct boot_params *params; @@ -108,9 +110,7 @@ void kdump_nmi_shootdown_cpus(void) /* Override the weak function in kernel/panic.c */ void crash_smp_send_stop(void) { - static int cpus_stopped; - - if (cpus_stopped) + if (crash_cpus_stopped) return; if (smp_ops.crash_stop_other_cpus) @@ -118,7 +118,7 @@ void crash_smp_send_stop(void) else smp_send_stop(); - cpus_stopped = 1; + crash_cpus_stopped = true; } #else diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c index fa700b46588e..2fc42b8402ac 100644 --- a/arch/x86/kernel/reboot.c +++ b/arch/x86/kernel/reboot.c @@ -589,8 +589,18 @@ static void native_machine_emergency_restart(void) int orig_reboot_type = reboot_type; unsigned short mode; - if (reboot_emergency) - emergency_vmx_disable_all(); + /* + * We can reach this point in the end of panic path, having + * NMI-disabled all secondary CPUs. This process involves + * disabling the CPU virtualization technologies, so if that + * is the case, we only miss disabling the local CPU VMX... + */ + if (reboot_emergency) { + if (!crash_cpus_stopped) + emergency_vmx_disable_all(); + else + cpu_emergency_vmxoff(); + } tboot_shutdown(TB_SHUTDOWN_REBOOT);
In the panic path we have a list of functions to be called, the panic notifiers - such callbacks perform various actions in the machine's last breath, and sometimes users want them to run before kdump. We have the parameter "crash_kexec_post_notifiers" for that. When such parameter is used, the function "crash_smp_send_stop()" is executed to poweroff all secondary CPUs through the NMI-shootdown mechanism; part of this process involves disabling virtualization features in all CPUs (except the main one). Now, in the emergency restart procedure we have also a way of disabling VMX in all CPUs, using the same NMI-shootdown mechanism; what happens though is that in case we already NMI-disabled all CPUs, the emergency restart fails due to a second addition of the same items in the NMI list, as per the following log output: sysrq: Trigger a crash Kernel panic - not syncing: sysrq triggered crash [...] Rebooting in 2 seconds.. list_add double add: new=<addr1>, prev=<addr2>, next=<addr1>. ------------[ cut here ]------------ kernel BUG at lib/list_debug.c:29! invalid opcode: 0000 [#1] PREEMPT SMP PTI In order to reproduce the problem, users just need to set the kernel parameter "crash_kexec_post_notifiers" *without* kdump set in any system with the VMX feature present. Since there is no benefit in re-disabling VMX in all CPUs in case it was already done, this patch prevents that by guarding the restart routine against doubly issuing NMIs unnecessarily. Notice we still need to disable VMX locally in the emergency restart. Fixes: ed72736183c4 ("x86/reboot: Force all cpus to exit VMX root if VMX is supported) Fixes: 0ee59413c967 ("x86/panic: replace smp_send_stop() with kdump friendly version in panic path") Cc: David P. Reed <dpreed@deepplum.com> Cc: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Sean Christopherson <seanjc@google.com> Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com> --- arch/x86/include/asm/cpu.h | 1 + arch/x86/kernel/crash.c | 8 ++++---- arch/x86/kernel/reboot.c | 14 ++++++++++++-- 3 files changed, 17 insertions(+), 6 deletions(-)