x86,perf: Disable intel_bts when PTI
Subsystems (2)
Peter Zijlstra
maintainerPATCH 1/1 x86,perf: Disable intel_bts when PTI
The intel_bts driver does not use the 'normal' BTS buffer we allocated space for in the cpu_entry_area but instead uses the memory allocated for the perf AUX buffer.
This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears.
For now disable this driver when PTI is enabled, we'll try and sort something out later.
--- arch/x86/events/intel/bts.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
arch/x86/events/intel/bts.c +18 -0
@@ -582,6 +582,24 @@ static __init int bts_init(void)if (!boot_cpu_has (X86_FEATURE_DTES64 ) || !x86_pmu.bts)return -ENODEV ; +if (boot_cpu_has (X86_FEATURE_PTI )) { +/* +* BTS hardware writes through a virtual memory map we must +* either use the kernel physical map, or the user mapping of +* the AUX buffer. +* +* However, since this driver supports per-CPU and per-task inherit +* we cannot use the user mapping since it will not be availble +* if we're not running the owning process. +* +* With PTI we can't use the kernal map either, because its not +* there when we run userspace. +* +* For now, disable this driver when using PTI. +*/ +return -ENODEV ; + } + bts_pmu.capabilities =PERF_PMU_CAP_AUX_NO_SG |PERF_PMU_CAP_ITRACE |PERF_PMU_CAP_EXCLUSIVE ; bts_pmu.task_ctx_nr = perf_sw_context;
Quoted 8 lines · Peter Zijlstra
The
intel_btsdriver does not use the 'normal' BTS buffer we allocated space for in thecpu_entry_areabut instead uses the memory allocated for the perf AUX buffer.This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears.
Quoted 7 lines
For now disable this driver when PTI is enabled, we'll try and sort something out later.
Cc Alexander Shishkin <[email protected]>Reported-by Vince Weaver <[email protected]>Reported-by Robert Święcki <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>
Can you add a:
Cc: stable <[email protected]>
please?
thanks,
greg k-h
Quoted 12 lines · Greg KH
Quoted 11 lines · Peter Zijlstra
The
intel_btsdriver does not use the 'normal' BTS buffer we allocated space for in thecpu_entry_areabut instead uses the memory allocated for the perf AUX buffer.This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears.
For now disable this driver when PTI is enabled, we'll try and sort something out later.
Quoted 8 lines
Quoted 4 lines
Cc Alexander Shishkin <[email protected]>Reported-by Vince Weaver <[email protected]>Reported-by Robert Święcki <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>Can you add a:
Cc: stable <[email protected]>
please?
Already in the commit ...
Quoted 12 lines · Greg KH
Quoted 11 lines · Peter Zijlstra
The
intel_btsdriver does not use the 'normal' BTS buffer we allocated space for in thecpu_entry_areabut instead uses the memory allocated for the perf AUX buffer.This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears.
For now disable this driver when PTI is enabled, we'll try and sort something out later.
Quoted 8 lines
Quoted 4 lines
Cc Alexander Shishkin <[email protected]>Reported-by Vince Weaver <[email protected]>Reported-by Robert Święcki <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>Can you add a:
Cc: stable <[email protected]>
please?
Note that these fixes are all going into tip:x86/pti, which is still a backporting friendly linear set of commits that should all be backported.
So once 99a9dc98ba52 goes upstream, all of v4.14..99a9dc98ba52 the remaining bits can go to -stable as well.
Will keep you updated about the latest status periodically, right now the latest that is already upstream is:
de791821c295: x86/pti: Rename BUG_CPU_INSECURE to BUG_CPU_MELTDOWN
i.e. all 194 commits in the v4.14..de791821c295 range should be queued up for -stable. (And most of them already are.)
There are 28 new commits in tip:x86/pti that will (hopefully) go upstream later today.
Thanks,
Ingo
Quoted 30 lines · Ingo Molnar
Quoted 12 lines · Greg KH
Quoted 11 lines · Peter Zijlstra
The
intel_btsdriver does not use the 'normal' BTS buffer we allocated space for in thecpu_entry_areabut instead uses the memory allocated for the perf AUX buffer.This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears.
For now disable this driver when PTI is enabled, we'll try and sort something out later.
Quoted 8 lines
Quoted 4 lines
Cc Alexander Shishkin <[email protected]>Reported-by Vince Weaver <[email protected]>Reported-by Robert Święcki <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>Can you add a:
Cc: stable <[email protected]>
please?Note that these fixes are all going into tip:x86/pti, which is still a backporting friendly linear set of commits that should all be backported.
So once
99a9dc98ba52goes upstream, all of v4.14..99a9dc98ba52 the remaining bits can go to -stable as well.
Quoted 7 lines
Will keep you updated about the latest status periodically, right now the latest that is already upstream is:
de791821c295: x86/pti: RenameBUG_CPU_INSECUREtoBUG_CPU_MELTDOWNi.e. all 194 commits in the v4.14..de791821c295 range should be queued up for -stable. (And most of them already are.)
I hope I got all of those in that branch already, as I've been watching it :)
Quoted 2 lines
There are 28 new commits in tip:x86/pti that will (hopefully) go upstream later today.
Great, again, thanks for doing it this way, it has made it much easier for me.
greg k-h
tip-bot for Peter Zijlstra
merge botx86,perf: Disable intel_bts when PTI
Commit-ID: 99a9dc98ba52267ce5e062b52de88ea1f1b2a7d8 Gitweb: https://git.kernel.org/tip/99a9dc98ba52267ce5e062b52de88ea1f1b2a7d8 Author: Peter Zijlstra <[email protected]> AuthorDate: Sun, 14 Jan 2018 11:27:13 +0100 Committer: Thomas Gleixner <[email protected]> CommitDate: Sun, 14 Jan 2018 11:42:10 +0100
x86,perf: Disable intel_bts when PTI
The intel_bts driver does not use the 'normal' BTS buffer which is exposed through the cpu_entry_area but instead uses the memory allocated for the perf AUX buffer.
This obviously comes apart when using PTI because then the kernel mapping; which includes that AUX buffer memory; disappears. Fixing this requires to expose a mapping which is visible in all context and that's not trivial.
As a quick fix disable this driver when PTI is enabled to prevent malfunction.
--- arch/x86/events/intel/bts.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)
arch/x86/events/intel/bts.c +18 -0
@@ -582,6 +582,24 @@ static __init int bts_init(void)if (!boot_cpu_has (X86_FEATURE_DTES64 ) || !x86_pmu.bts)return -ENODEV ; +if (boot_cpu_has (X86_FEATURE_PTI )) { +/* +* BTS hardware writes through a virtual memory map we must +* either use the kernel physical map, or the user mapping of +* the AUX buffer. +* +* However, since this driver supports per-CPU and per-task inherit +* we cannot use the user mapping since it will not be availble +* if we're not running the owning process. +* +* With PTI we can't use the kernal map either, because its not +* there when we run userspace. +* +* For now, disable this driver when using PTI. +*/ +return -ENODEV ; + } + bts_pmu.capabilities =PERF_PMU_CAP_AUX_NO_SG |PERF_PMU_CAP_ITRACE |PERF_PMU_CAP_EXCLUSIVE ; bts_pmu.task_ctx_nr = perf_sw_context;