[PATCH v9 00/27] x86: load FPU registers on return to userland

From: Sebastian Andrzej Siewior
Date: Wed Apr 03 2019 - 12:45:26 EST


This is a refurbished series originally started by by Rik van Riel. The
goal is load the FPU registers on return to userland and not on every
context switch. By this optimisation we can:
- avoid loading the registers if the task stays in kernel and does
not return to userland
- make kernel_fpu_begin() cheaper: it only saves the registers on the
first invocation. The second invocation does not need save them again.

To access the FPU registers in kernel we need:
- disable preemption to avoid that the scheduler switches tasks. By
doing so it would set TIF_NEED_FPU_LOAD and the FPU registers would be
not valid.
- disable BH because the softirq might use kernel_fpu_begin() and then
set TIF_NEED_FPU_LOAD instead loading the FPU registers on completion.

v8âv9:
Rebased on top of v5.1-rc3 plus:
- using the proper register pointer in __fpu__restore_sig() for the
fsave case.

- dropped a WARN_ON() from switch_fpu_finish() if the PKRU state is
missing. On a preemptible kernel this warning will trigger if the
task gets preempted in fpstate_init() during memset() and
fpstate_init_xstate().

- Adding a fastpath to __fpu__restore_sig() and
copy_fpstate_to_sigframe() save/restore FPU state to/from userland.
Thomas Gleixner asked for this. The series first does the slowpath and
then the last few patches add/enable the fast path. This ensures that
the slow works correctly (during a bisect) and then it should be
rarely used.

Some numbers generated by
taskset 2 lat_sig -P 1 -W 64 -N 5000 catch

v5.1-rc3
catch 64
Signal handler overhead: 1.7531 microseconds
Signal handler overhead: 1.7547 microseconds
Signal handler overhead: 1.7362 microseconds
catch 32
Signal handler overhead: 2.4044 microseconds
Signal handler overhead: 2.3732 microseconds
Signal handler overhead: 2.3979 microseconds

up to x86/fpu: Defer FPU state load until return to userspace
catch 64
Signal handler overhead: 2.7556 microseconds
Signal handler overhead: 2.7878 microseconds
Signal handler overhead: 2.7664 microseconds
catch 32
Signal handler overhead: 2.5961 microseconds
Signal handler overhead: 2.5916 microseconds
Signal handler overhead: 2.6091 microseconds

up to x86/fpu: Add a fastpath to copy_fpstate_to_sigframe()
catch 64
Signal handler overhead: 1.8573 microseconds
Signal handler overhead: 1.8469 microseconds
Signal handler overhead: 1.8554 microseconds
catch 32
Signal handler overhead: 2.4875 microseconds
Signal handler overhead: 2.4822 microseconds
Signal handler overhead: 2.4806 microseconds
----
complete series
catch 64
Signal handler overhead: 1.8463 microseconds
Signal handler overhead: 1.8431 microseconds
Signal handler overhead: 1.8609 microseconds
catch 32
Signal handler overhead: 2.4907 microseconds
Signal handler overhead: 2.4771 microseconds
Signal handler overhead: 2.4939 microseconds

v7âv8:
Rebased on top of v5.1-rc1. And then:
- Remove a WARN_ON() in switch_fpu_finish. Turns out it can trigger
during a preemption while the xstate is initialized.

- Provide __read_pkru_ins() and __write_pkru_ins which are symmetrical.
__write_pkru() does the "write only if the value is different from
current" check.

- The kernel threads now load `init_pkru_value' instead of `0' for the
PKRU value.

- The PKRU value is also written into its xsave area of init_fpstate.

v6âv7:
Rebased on top of v5.0-rc7 and addressed Borislav Petkov's review.

v5âv6:
Rebased on top of v5.0-rc1. Integrated a few fixes which I noticed while
looking over the patches, dropped the first few patches which were
already applied.

v4âv5:
Rebased on top of a fix, noticed a problem with XSAVES and then redid
the restore on sig return (patch #26 to #28).

I don't like very much the sig save+restore thing that we are doing. It
has been always like that. I *think* that this is just because we have
nowhere to stash the FPU state while we are handling the signal. We
could add another fpu->state for the signal handler and avoid the thing.
Debian code-search revealed that `criu' is using it (and I didn't
figure out why). Nothing else (that is packaged in Debian). Maybe we
could get rid of this and if `criu' would then use a dedicated interface
for its needs rather the signal interface that happen to do what it
wants :)

v3âv4:
It has been suggested to remove the `initialized' member of the struct
fpu because it should not required be needed with lazy-FPU-restore and
would make the review easier. This is the first part of the series, the
second is basically the rebase of the v3 queue. As a result, the
diffstat became negative (which wasn't the case in previous version) :)
I tried to incorporate all the review comments that came up, some of
them were "outdated" after the removal of the `initialized' member. I'm
sorry should I missed any.

v1âv3:
v2 was never posted. I followed the idea to completely decouple PKRU
from xstate. This didn't quite work and made a few things complicated.
One obvious required fixup is copy_fpstate_to_sigframe() where the PKRU
state needs to be fiddled into xstate. This required another
xfeatures_mask so that the sanity checks were performed and
xstate_offsets would be computed. Additionally ptrace also reads/sets
xstate in order to get/set the register and PKRU is one of them. So this
would need some fiddle, too.
In v3 I dropped that decouple idea. I also learned that the wrpkru
instruction is not privileged and so caching it in kernel does not work.
Instead I keep PKRU in xstate area and load it at context switch time
while the remaining registers are deferred (until return to userland).
The offset of PKRU within xstate is enumerated at boot time so why not
use it.

The following changes since commit 79a3aaa7b82e3106be97842dedfd8429248896e6:

Linux 5.1-rc3 (2019-03-31 14:39:29 -0700)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/bigeasy/staging.git x86_fpu_rtu_v9

for you to fetch changes up to c5265b05d167b3be19c787ebb913dc138485fa25:

x86/pkeys: add PKRU value to init_fpstate (2019-04-03 18:19:10 +0200)

----------------------------------------------------------------
Rik van Riel (5):
x86/fpu: Add (__)make_fpregs_active helpers
x86/fpu: Eager switch PKRU state
x86/fpu: Always store the registers in copy_fpstate_to_sigframe()
x86/fpu: Prepare copy_fpstate_to_sigframe() for TIF_NEED_FPU_LOAD
x86/fpu: Defer FPU state load until return to userspace

Sebastian Andrzej Siewior (22):
x86/fpu: Remove fpu->initialized usage in __fpu__restore_sig()
x86/fpu: Remove fpu__restore()
x86/fpu: Remove preempt_disable() in fpu__clear()
x86/fpu: Always init the `state' in fpu__clear()
x86/fpu: Remove fpu->initialized usage in copy_fpstate_to_sigframe()
x86/fpu: Don't save fxregs for ia32 frames in copy_fpstate_to_sigframe()
x86/fpu: Remove fpu->initialized
x86/fpu: Remove user_fpu_begin()
x86/fpu: Make __raw_xsave_addr() use feature number instead of mask
x86/fpu: Make get_xsave_field_ptr() and get_xsave_addr() use feature number instead of mask
x86/pkru: Provide .*_pkru_ins() functions
x86/fpu: Only write PKRU if it is different from current
x86/pkeys: Don't check if PKRU is zero before writting it
x86/entry: Add TIF_NEED_FPU_LOAD
x86/fpu: Update xstate's PKRU value on write_pkru()
x86/fpu: Inline copy_user_to_fpregs_zeroing()
x86/fpu: Let __fpu__restore_sig() restore the !32bit+fxsr frame from kernel memory
x86/fpu: Merge the two code paths in __fpu__restore_sig()
x86/fpu: Add a fastpath to __fpu__restore_sig()
x86/fpu: Add a fastpath to copy_fpstate_to_sigframe()
x86/fpu: Restore FPU register in copy_fpstate_to_sigframe() in order to use the fastpath
x86/pkeys: add PKRU value to init_fpstate

Documentation/preempt-locking.txt | 1 -
arch/x86/entry/common.c | 8 ++
arch/x86/ia32/ia32_signal.c | 17 ++-
arch/x86/include/asm/fpu/api.h | 31 ++++++
arch/x86/include/asm/fpu/internal.h | 130 ++++++++++++++++------
arch/x86/include/asm/fpu/signal.h | 2 +-
arch/x86/include/asm/fpu/types.h | 9 --
arch/x86/include/asm/fpu/xstate.h | 5 +-
arch/x86/include/asm/pgtable.h | 28 ++++-
arch/x86/include/asm/special_insns.h | 18 +++-
arch/x86/include/asm/thread_info.h | 2 +
arch/x86/include/asm/trace/fpu.h | 8 +-
arch/x86/kernel/cpu/common.c | 5 +
arch/x86/kernel/fpu/core.c | 194 ++++++++++++++++-----------------
arch/x86/kernel/fpu/init.c | 2 -
arch/x86/kernel/fpu/regset.c | 24 ++---
arch/x86/kernel/fpu/signal.c | 202 +++++++++++++++++++++--------------
arch/x86/kernel/fpu/xstate.c | 42 ++++----
arch/x86/kernel/process.c | 2 +-
arch/x86/kernel/process_32.c | 11 +-
arch/x86/kernel/process_64.c | 11 +-
arch/x86/kernel/signal.c | 17 ++-
arch/x86/kernel/traps.c | 2 +-
arch/x86/kvm/vmx/vmx.c | 2 +-
arch/x86/kvm/x86.c | 48 +++++----
arch/x86/math-emu/fpu_entry.c | 3 -
arch/x86/mm/mpx.c | 6 +-
arch/x86/mm/pkeys.c | 21 ++--
28 files changed, 496 insertions(+), 355 deletions(-)

Sebastian