xen/vcpu: ignore VCPU_SSHOTTMR_future
PATCH v3 1/1 xen/vcpu: ignore VCPU_SSHOTTMR_future
The usage of VCPU_SSHOTTMR_future in Linux prior to 4.7 is bogus. When the hypervisor returns -ETIME (timeout in the past) Linux keeps retrying to setup the timer with a higher timeout instead of self-injecting a timer interrupt.
On boxes without any hardware assistance for logdirty we have seen HVM Linux guests < 4.7 with 32vCPUs give up trying to setup the timer when logdirty is enabled:
CE: Reprogramming failure. Giving up
CE: xen increased min_delta_ns to 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: xen increased min_delta_ns to 506250 nsec
CE: xen increased min_delta_ns to 759375 nsec
CE: xen increased min_delta_ns to 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
Freezing user space processes ...
INFO: rcu_sched detected stalls on CPUs/tasks: { 14} (detected by 10, t=60002 jiffies, g=4006, c=4005, q=14130)
Task dump for CPU 14:
swapper/14 R running task 0 0 1 0x00000000
Call Trace:
[<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0
[<ffffffff907b9bde>] ? default_idle+0x1e/0xd0
[<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0
[<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0
[<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270
[<ffffffff900000d5>] ? start_cpu+0x5/0x14
INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=6922, c=6921, q=7013)
Task dump for CPU 26:
swapper/26 R running task 0 0 1 0x00000000
Call Trace:
[<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0
[<ffffffff907b9bde>] ? default_idle+0x1e/0xd0
[<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0
[<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0
[<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270
[<ffffffff900000d5>] ? start_cpu+0x5/0x14
INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=8499, c=8498, q=7664)
Task dump for CPU 26:
swapper/26 R running task 0 0 1 0x00000000
Call Trace:
[<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0
[<ffffffff907b9bde>] ? default_idle+0x1e/0xd0
[<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0
[<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0
[<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270
[<ffffffff900000d5>] ? start_cpu+0x5/0x14Thus leading to CPU stalls and a broken system as a result.
Workaround this bogus usage by ignoring the VCPU_SSHOTTMR_future in the hypervisor. Old Linux versions are the only ones known to have (wrongly) attempted to use the flag, and ignoring it is compatible with the behavior expected by any guests setting that flag.
Note the usage of the flag has been removed from Linux by commit:
c06b6d70feb3 xen/x86: don't lose event interrupts
Which landed in Linux 4.7.
---
Changes since v2:
- Reword the comment about VCPU_SSHOTTMR_future being ignored.
- s/ENOTIME/ETIME/ in the commit message.
- Reword changelog message.
Changes since v1:
- Just ignore the flag, as there's no ABI breakage.
--- CHANGELOG.md | 2 ++ xen/common/domain.c | 13 ++++++++++--- xen/include/public/vcpu.h | 5 ++++- 3 files changed, 16 insertions(+), 4 deletions(-)
CHANGELOG.md +2 -0
@@ -9,6 +9,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) ### Changed - Repurpose command line gnttab_max_{maptrack_,}frames options so they don't cap toolstack provided values. + - Ignore VCPUOP_set_singleshot_timer's VCPU_SSHOTTMR_future flag. The only + known user doesn't use it properly, leading to in-guest breakage. ### Added - On x86, support for features new in Intel Sapphire Rapids CPUs:
xen/common/domain.c +10 -3
@@ -1762,9 +1762,16 @@ long common_vcpu_op(int cmd, struct vcpu *v, XEN_GUEST_HANDLE_PARAM(void) arg)if (copy_from_guest (&set, arg,1 ) )return -EFAULT ; -if ( (set.flags & VCPU_SSHOTTMR_future) && - (set.timeout_abs_ns <NOW ()) ) -return -ETIME ; +if ( set.timeout_abs_ns <NOW () ) + { +/* +* Simplify the logic if the timeout has already expired and just +* inject the event. +*/ +stop_timer (&v->singleshot_timer); +send_timer_event (v); +break ; + }migrate_timer (&v->singleshot_timer,smp_processor_id ());set_timer (&v->singleshot_timer, set.timeout_abs_ns);
xen/include/public/vcpu.h +4 -1
@@ -150,7 +150,10 @@ typedef struct vcpu_set_singleshot_timer vcpu_set_singleshot_timer_t;DEFINE_XEN_GUEST_HANDLE (vcpu_set_singleshot_timer_t);/* Flags to VCPUOP_set_singleshot_timer. */ -/* Require the timeout to be in the future (return -ETIME if it's passed). */ +/* +* Request the timeout to be in the future (return -ETIME if it's passed) +* but can be ignored by the hypervisor. +*/ #define _VCPU_SSHOTTMR_future (0)#define VCPU_SSHOTTMR_future (1U << _VCPU_SSHOTTMR_future)
Signature · 1 line
2.40.0
Quoted 58 lines · Roger Pau Monne
The usage of
VCPU_SSHOTTMR_futurein Linux prior to 4.7 is bogus. When the hypervisor returns-ETIME(timeout in the past) Linux keeps retrying to setup the timer with a higher timeout instead of self-injecting a timer interrupt.On boxes without any hardware assistance for logdirty we have seen HVM Linux guests < 4.7 with 32vCPUs give up trying to setup the timer when logdirty is enabled:
CE: Reprogramming failure. Giving up
CE: xen increasedmin_delta_nsto 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: xen increasedmin_delta_nsto 506250 nsec
CE: xen increasedmin_delta_nsto 759375 nsec
CE: xen increasedmin_delta_nsto 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
Freezing user space processes ...INFO: rcu_sched detected stalls on CPUs/tasks: { 14} (detected by 10, t=60002 jiffies, g=4006, c=4005, q=14130) Task dump for CPU 14: swapper/14 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14 INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=6922, c=6921, q=7013) Task dump for CPU 26: swapper/26 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14 INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=8499, c=8498, q=7664) Task dump for CPU 26: swapper/26 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14Thus leading to CPU stalls and a broken system as a result.
Workaround this bogus usage by ignoring the
VCPU_SSHOTTMR_futurein the hypervisor. Old Linux versions are the only ones known to have (wrongly) attempted to use the flag, and ignoring it is compatible with the behavior expected by any guests setting that flag.
Quoted 8 lines
Note the usage of the flag has been removed from Linux by commit:
c06b6d70feb3xen/x86: don't lose event interruptsWhich landed in Linux 4.7.
Signed-off-by Roger Pau Monné <[email protected]>
A little hesitantly, but since no-one else appears to show any interest:
Jan
Quoted 63 lines · Jan Beulich
Quoted 62 lines · Roger Pau Monne
The usage of
VCPU_SSHOTTMR_futurein Linux prior to 4.7 is bogus. When the hypervisor returns-ETIME(timeout in the past) Linux keeps retrying to setup the timer with a higher timeout instead of self-injecting a timer interrupt.On boxes without any hardware assistance for logdirty we have seen HVM Linux guests < 4.7 with 32vCPUs give up trying to setup the timer when logdirty is enabled:
CE: Reprogramming failure. Giving up
CE: xen increasedmin_delta_nsto 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: xen increasedmin_delta_nsto 506250 nsec
CE: xen increasedmin_delta_nsto 759375 nsec
CE: xen increasedmin_delta_nsto 1000000 nsec
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
CE: Reprogramming failure. Giving up
Freezing user space processes ...INFO: rcu_sched detected stalls on CPUs/tasks: { 14} (detected by 10, t=60002 jiffies, g=4006, c=4005, q=14130) Task dump for CPU 14: swapper/14 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14 INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=6922, c=6921, q=7013) Task dump for CPU 26: swapper/26 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14 INFO: rcu_sched detected stalls on CPUs/tasks: { 26} (detected by 24, t=60002 jiffies, g=8499, c=8498, q=7664) Task dump for CPU 26: swapper/26 R running task 0 0 1 0x00000000 Call Trace: [<ffffffff90160f5d>] ? rcu_eqs_enter_common.isra.30+0x3d/0xf0 [<ffffffff907b9bde>] ? default_idle+0x1e/0xd0 [<ffffffff90039570>] ? arch_cpu_idle+0x20/0xc0 [<ffffffff9010820a>] ? cpu_startup_entry+0x14a/0x1e0 [<ffffffff9005d3a7>] ? start_secondary+0x1f7/0x270 [<ffffffff900000d5>] ? start_cpu+0x5/0x14Thus leading to CPU stalls and a broken system as a result.
Workaround this bogus usage by ignoring the
VCPU_SSHOTTMR_futurein the hypervisor. Old Linux versions are the only ones known to have (wrongly) attempted to use the flag, and ignoring it is compatible with the behavior expected by any guests setting that flag.Note the usage of the flag has been removed from Linux by commit:
c06b6d70feb3xen/x86: don't lose event interrupts
Quoted 7 lines
Quoted 4 lines
Which landed in Linux 4.7.
Signed-off-by Roger Pau Monné <[email protected]>A little hesitantly, but since no-one else appears to show any interest:
Acked-by Jan Beulich <[email protected]>
Thanks.
Roger.