public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: "H.J. Lu" <hjl.tools@gmail.com>
To: DJ Delorie <dj@redhat.com>
Cc: GNU C Library <libc-alpha@sourceware.org>
Subject: [PATCH v2] x86: Disable OSXSAVE related features [BZ #27605]
Date: Fri, 19 Mar 2021 15:22:14 -0700	[thread overview]
Message-ID: <CAMe9rOoTXnQDxSzmL=7Mvj89Ynx-PSUeGZFthYCoQeB2zeWhaA@mail.gmail.com> (raw)
In-Reply-To: <xn7dm2n7lw.fsf@rhel8.vm>

[-- Attachment #1: Type: text/plain, Size: 451 bytes --]

On Fri, Mar 19, 2021 at 2:42 PM DJ Delorie <dj@redhat.com> wrote:
>
> "H.J. Lu" <hjl.tools@gmail.com> writes:
> > If OSXSAVE is disabled by glibc tunables, disable features which depend
> > on OSXSAVE.
>
> This one seems to ignore -XSAVE, both on an 11th gen (-XSAVE,-XSAVEC ->
> xsave()) and on 4th (just -XSAVE).  Is this patch supposed to be used
> with the previous patch, or standalone?
>

Fixed.  Here is the v2 patch.  OK for master?

-- 
H.J.

[-- Attachment #2: v2-0001-x86-Disable-OSXSAVE-related-features-BZ-27605.patch --]
[-- Type: text/x-patch, Size: 3718 bytes --]

From 36e3a13647d3915d307281334d1cba13ab975a1a Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Fri, 19 Mar 2021 06:15:37 -0700
Subject: [PATCH v2] x86: Disable OSXSAVE related features [BZ #27605]

1. Support GLIBC_TUNABLES=glibc.cpu.hwcaps=-XSAVE.
2. If OSXSAVE is disabled by glibc tunables, disable features which
depend on OSXSAVE.
3. Clear xsave_state_size if both XSAVE and XSAVEC aren't usable.
---
 sysdeps/x86/cpu-features.c | 46 ++++++++++++++++++++++++++++++++++++++
 sysdeps/x86/cpu-tunables.c |  1 +
 2 files changed, 47 insertions(+)

diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
index d7248cbb45..a9c62b4231 100644
--- a/sysdeps/x86/cpu-features.c
+++ b/sysdeps/x86/cpu-features.c
@@ -654,6 +654,52 @@ no_cpuid:
 
 #if HAVE_TUNABLES
   TUNABLE_GET (hwcaps, tunable_val_t *, TUNABLE_CALLBACK (set_hwcaps));
+
+  if (!CPU_FEATURE_USABLE_P (cpu_features, OSXSAVE))
+    {
+      /* These features are usable only if OSXSAVE is usable.  */
+      CPU_FEATURE_UNSET (cpu_features, AVX);
+      CPU_FEATURE_UNSET (cpu_features, AVX2);
+      CPU_FEATURE_UNSET (cpu_features, AVX_VNNI);
+      CPU_FEATURE_UNSET (cpu_features, FMA);
+      CPU_FEATURE_UNSET (cpu_features, VAES);
+      CPU_FEATURE_UNSET (cpu_features, VPCLMULQDQ);
+      CPU_FEATURE_UNSET (cpu_features, XOP);
+      CPU_FEATURE_UNSET (cpu_features, F16C);
+      CPU_FEATURE_UNSET (cpu_features, AVX512F);
+      CPU_FEATURE_UNSET (cpu_features, AVX512CD);
+      CPU_FEATURE_UNSET (cpu_features, AVX512ER);
+      CPU_FEATURE_UNSET (cpu_features, AVX512PF);
+      CPU_FEATURE_UNSET (cpu_features, AVX512VL);
+      CPU_FEATURE_UNSET (cpu_features, AVX512DQ);
+      CPU_FEATURE_UNSET (cpu_features, AVX512BW);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_4FMAPS);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_4VNNIW);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_BITALG);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_IFMA);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_VBMI);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_VBMI2);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_VNNI);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_VPOPCNTDQ);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_VP2INTERSECT);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_BF16);
+      CPU_FEATURE_UNSET (cpu_features, AVX512_FP16);
+      CPU_FEATURE_UNSET (cpu_features, AMX_BF16);
+      CPU_FEATURE_UNSET (cpu_features, AMX_TILE);
+      CPU_FEATURE_UNSET (cpu_features, AMX_INT8);
+      CPU_FEATURE_UNSET (cpu_features, XSAVE);
+      CPU_FEATURE_UNSET (cpu_features, XSAVEOPT);
+      CPU_FEATURE_UNSET (cpu_features, XSAVEC);
+      CPU_FEATURE_UNSET (cpu_features, XGETBV_ECX_1);
+      CPU_FEATURE_UNSET (cpu_features, XFD);
+
+      CPU_FEATURE_UNSET (cpu_features, FMA4);
+    }
+
+  /* Clear xsave_state_size if both XSAVE and XSAVEC aren't usable.  */
+  if (!CPU_FEATURE_USABLE_P (cpu_features, XSAVE)
+      && !CPU_FEATURE_USABLE_P (cpu_features, XSAVEC))
+    cpu_features->xsave_state_size = 0;
 #elif defined SHARED
   /* Reuse dl_platform, dl_hwcap and dl_hwcap_mask for x86.  The
      glibc.cpu.hwcap_mask tunable is initialized already, so no
diff --git a/sysdeps/x86/cpu-tunables.c b/sysdeps/x86/cpu-tunables.c
index 126896f41b..61b05e5b1d 100644
--- a/sysdeps/x86/cpu-tunables.c
+++ b/sysdeps/x86/cpu-tunables.c
@@ -168,6 +168,7 @@ TUNABLE_CALLBACK (set_hwcaps) (tunable_val_t *valp)
 	      CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, MOVBE, 5);
 	      CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, SHSTK, 5);
 	      CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, SSSE3, 5);
+	      CHECK_GLIBC_IFUNC_CPU_OFF (n, cpu_features, XSAVE, 5);
 	    }
 	  break;
 	case 6:
-- 
2.30.2


  reply	other threads:[~2021-03-19 22:22 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-19 13:27 [PATCH] " H.J. Lu
2021-03-19 21:42 ` DJ Delorie
2021-03-19 22:22   ` H.J. Lu [this message]
2021-03-20  0:06     ` [PATCH v3] x86: Properly disable XSAVE " H.J. Lu
2021-03-22 18:32       ` DJ Delorie
2021-03-22 18:38         ` H.J. Lu
2021-03-22 18:42           ` DJ Delorie
2021-03-23 13:08             ` H.J. Lu

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='CAMe9rOoTXnQDxSzmL=7Mvj89Ynx-PSUeGZFthYCoQeB2zeWhaA@mail.gmail.com' \
    --to=hjl.tools@gmail.com \
    --cc=dj@redhat.com \
    --cc=libc-alpha@sourceware.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 read-only IMAP folder(s) and NNTP newsgroup(s).