kvm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: linux-kernel@vger.kernel.org
Cc: x86@kernel.org, "Andy Lutomirski" <luto@kernel.org>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Radim Krčmář" <rkrcmar@redhat.com>,
	kvm@vger.kernel.org, "Jason A. Donenfeld" <Jason@zx2c4.com>,
	"Rik van Riel" <riel@surriel.com>,
	"Dave Hansen" <dave.hansen@linux.intel.com>,
	"Sebastian Andrzej Siewior" <bigeasy@linutronix.de>
Subject: [PATCH 26/27] x86/fpu: Restore FPU register in copy_fpstate_to_sigframe() in order to use the fastpath
Date: Wed,  3 Apr 2019 18:41:55 +0200	[thread overview]
Message-ID: <20190403164156.19645-27-bigeasy@linutronix.de> (raw)
In-Reply-To: <20190403164156.19645-1-bigeasy@linutronix.de>

If a task is scheduled out and receives a signal then it won't be able to take
the fastpath because the register aren't available. The slowpath is more
expensive compared to xrstor + xsave which usually succeeds.

Some clock_gettime() numbers from a bigger box with AVX512 during bootup:
- __fpregs_load_activate() takes 140ns - 350ns. If it was the most recent FPU
  context on the CPU then the optimisation in __fpregs_load_activate() will
  skip the load (which was disabled during the test).

- copy_fpregs_to_sigframe() takes 200ns - 450ns if it succeeds. On a
  pagefault it is 1.8us - 3us usually in the 2.6us area.

- The slowpath takes 1.5 - 6us. Usually in the 2.6us area.

My testcases (including lat_sig) take the fastpath without
__fpregs_load_activate(). I expect this to be the majority.

Since the slowpath is in the >1us area it makes sense to load the
registers and attempt to save them directly. The direct save may fail
but should only happen on the first invocation or after fork() while the
page is RO.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 arch/x86/kernel/fpu/signal.c | 25 +++++++++++++------------
 1 file changed, 13 insertions(+), 12 deletions(-)

diff --git a/arch/x86/kernel/fpu/signal.c b/arch/x86/kernel/fpu/signal.c
index baf1588d7060c..16f700d5b3a47 100644
--- a/arch/x86/kernel/fpu/signal.c
+++ b/arch/x86/kernel/fpu/signal.c
@@ -176,19 +176,20 @@ int copy_fpstate_to_sigframe(void __user *buf, void __user *buf_fx, int size)
 
 	fpregs_lock();
 	/*
-	 * If we do not need to load the FPU registers at return to userspace
-	 * then the CPU has the current state. Try to save it directly to
-	 * userland's stack frame if it does not cause a pagefault. If it does,
-	 * try the slowpath.
+	 * Load the FPU register if they are not valid for the current task.
+	 * With a valid FPU state we can attempt to save the state directly to
+	 * userland's stack frame which will likely succeed. If it does not, do
+	 * the slowpath.
 	 */
-	if (!test_thread_flag(TIF_NEED_FPU_LOAD)) {
-		pagefault_disable();
-		ret = copy_fpregs_to_sigframe(buf_fx);
-		pagefault_enable();
-		if (ret)
-			copy_fpregs_to_fpstate(fpu);
-		set_thread_flag(TIF_NEED_FPU_LOAD);
-	}
+	if (test_thread_flag(TIF_NEED_FPU_LOAD))
+		__fpregs_load_activate();
+
+	pagefault_disable();
+	ret = copy_fpregs_to_sigframe(buf_fx);
+	pagefault_enable();
+	if (ret && !test_thread_flag(TIF_NEED_FPU_LOAD))
+		copy_fpregs_to_fpstate(fpu);
+	set_thread_flag(TIF_NEED_FPU_LOAD);
 	fpregs_unlock();
 
 	if (ret) {
-- 
2.20.1

  parent reply	other threads:[~2019-04-03 16:41 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-04-03 16:41 [PATCH v9 00/27] x86: load FPU registers on return to userland Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 01/27] x86/fpu: Remove fpu->initialized usage in __fpu__restore_sig() Sebastian Andrzej Siewior
2019-04-13 20:46   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 02/27] x86/fpu: Remove fpu__restore() Sebastian Andrzej Siewior
2019-04-13 20:47   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 03/27] x86/fpu: Remove preempt_disable() in fpu__clear() Sebastian Andrzej Siewior
2019-04-13 20:48   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 04/27] x86/fpu: Always init the `state' " Sebastian Andrzej Siewior
2019-04-13 20:48   ` [tip:x86/fpu] x86/fpu: Always init the state " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 05/27] x86/fpu: Remove fpu->initialized usage in copy_fpstate_to_sigframe() Sebastian Andrzej Siewior
2019-04-13 20:49   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 06/27] x86/fpu: Don't save fxregs for ia32 frames " Sebastian Andrzej Siewior
2019-04-13 20:50   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 07/27] x86/fpu: Remove fpu->initialized Sebastian Andrzej Siewior
2019-04-13 20:50   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 08/27] x86/fpu: Remove user_fpu_begin() Sebastian Andrzej Siewior
2019-04-13 20:51   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 09/27] x86/fpu: Add (__)make_fpregs_active helpers Sebastian Andrzej Siewior
2019-04-13 20:52   ` [tip:x86/fpu] x86/fpu: Add an __fpregs_load_activate() internal helper tip-bot for Rik van Riel
2019-04-03 16:41 ` [PATCH 10/27] x86/fpu: Make __raw_xsave_addr() use feature number instead of mask Sebastian Andrzej Siewior
2019-04-13 20:52   ` [tip:x86/fpu] x86/fpu: Make __raw_xsave_addr() use a " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 11/27] x86/fpu: Make get_xsave_field_ptr() and get_xsave_addr() use " Sebastian Andrzej Siewior
2019-04-13 20:53   ` [tip:x86/fpu] x86/fpu: Use a feature number instead of mask in two more helpers tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 12/27] x86/pkru: Provide .*_pkru_ins() functions Sebastian Andrzej Siewior
2019-04-10 16:36   ` Borislav Petkov
2019-04-10 16:52     ` Borislav Petkov
2019-04-10 21:25       ` Sebastian Andrzej Siewior
2019-04-10 21:29         ` Dave Hansen
2019-04-11 13:24           ` Borislav Petkov
2019-04-13 20:54   ` [tip:x86/fpu] x86/pkeys: Provide *pkru() helpers tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 13/27] x86/fpu: Only write PKRU if it is different from current Sebastian Andrzej Siewior
2019-04-13 20:55   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 14/27] x86/pkeys: Don't check if PKRU is zero before writting it Sebastian Andrzej Siewior
2019-04-13 20:55   ` [tip:x86/fpu] x86/pkeys: Don't check if PKRU is zero before writing it tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 15/27] x86/fpu: Eager switch PKRU state Sebastian Andrzej Siewior
2019-04-13 20:56   ` [tip:x86/fpu] " tip-bot for Rik van Riel
2019-04-03 16:41 ` [PATCH 16/27] x86/entry: Add TIF_NEED_FPU_LOAD Sebastian Andrzej Siewior
2019-04-13 20:57   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 17/27] x86/fpu: Always store the registers in copy_fpstate_to_sigframe() Sebastian Andrzej Siewior
2019-04-13 20:57   ` [tip:x86/fpu] " tip-bot for Rik van Riel
2019-04-03 16:41 ` [PATCH 18/27] x86/fpu: Prepare copy_fpstate_to_sigframe() for TIF_NEED_FPU_LOAD Sebastian Andrzej Siewior
2019-04-13 20:58   ` [tip:x86/fpu] " tip-bot for Rik van Riel
2019-04-03 16:41 ` [PATCH 19/27] x86/fpu: Update xstate's PKRU value on write_pkru() Sebastian Andrzej Siewior
2019-04-08 18:14   ` Dave Hansen
2019-04-08 20:03     ` Sebastian Andrzej Siewior
2019-04-13 20:59   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 20/27] x86/fpu: Inline copy_user_to_fpregs_zeroing() Sebastian Andrzej Siewior
2019-04-13 21:00   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 21/27] x86/fpu: Let __fpu__restore_sig() restore the !32bit+fxsr frame from kernel memory Sebastian Andrzej Siewior
2019-04-13 21:00   ` [tip:x86/fpu] x86/fpu: Restore from kernel memory on the 64-bit path too tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 22/27] x86/fpu: Merge the two code paths in __fpu__restore_sig() Sebastian Andrzej Siewior
2019-04-13 21:01   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 23/27] x86/fpu: Defer FPU state load until return to userspace Sebastian Andrzej Siewior
2019-04-12 14:36   ` Borislav Petkov
2019-04-12 15:24     ` Sebastian Andrzej Siewior
2019-04-12 16:22       ` Borislav Petkov
2019-04-12 16:37         ` Sebastian Andrzej Siewior
2019-04-12 16:48           ` Borislav Petkov
2019-04-12 17:19             ` Sebastian Andrzej Siewior
2019-04-12 17:29               ` Borislav Petkov
2019-04-15  9:14                 ` Sebastian Andrzej Siewior
2019-04-13 21:02   ` [tip:x86/fpu] " tip-bot for Rik van Riel
2019-04-03 16:41 ` [PATCH 24/27] x86/fpu: Add a fastpath to __fpu__restore_sig() Sebastian Andrzej Siewior
2019-04-08 17:05   ` Thomas Gleixner
2019-04-08 20:02     ` Sebastian Andrzej Siewior
2019-04-09  7:27       ` Thomas Gleixner
2019-04-12 17:17   ` Borislav Petkov
2019-04-12 17:27     ` Sebastian Andrzej Siewior
2019-04-13 21:02   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 25/27] x86/fpu: Add a fastpath to copy_fpstate_to_sigframe() Sebastian Andrzej Siewior
2019-04-13 21:03   ` [tip:x86/fpu] " tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` Sebastian Andrzej Siewior [this message]
2019-04-13 21:04   ` [tip:x86/fpu] x86/fpu: Restore regs in copy_fpstate_to_sigframe() in order to use the fastpath tip-bot for Sebastian Andrzej Siewior
2019-04-03 16:41 ` [PATCH 27/27] x86/pkeys: add PKRU value to init_fpstate Sebastian Andrzej Siewior
2019-04-13 21:05   ` [tip:x86/fpu] x86/pkeys: Add " tip-bot for Sebastian Andrzej Siewior
2019-04-04 14:01 ` [PATCH v9 00/27] x86: load FPU registers on return to userland David Laight
2019-04-04 14:14   ` 'Sebastian Andrzej Siewior'
2019-04-04 14:26     ` Andy Lutomirski
2019-04-04 14:31       ` Sebastian Andrzej Siewior
2019-04-04 15:10       ` David Laight
2019-04-08 17:08 ` Thomas Gleixner
2019-04-12 18:30 ` Borislav Petkov
2019-04-15  8:58   ` Sebastian Andrzej Siewior

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190403164156.19645-27-bigeasy@linutronix.de \
    --to=bigeasy@linutronix.de \
    --cc=Jason@zx2c4.com \
    --cc=dave.hansen@linux.intel.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luto@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=riel@surriel.com \
    --cc=rkrcmar@redhat.com \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).