From: Ingo Molnar <mingo@kernel.org>
To: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo2.kernel.org@gmail.com>,
linux-kernel@vger.kernel.org,
Andy Lutomirski <luto@amacapital.net>,
Borislav Petkov <bp@alien8.de>, Fenghua Yu <fenghua.yu@intel.com>,
"H. Peter Anvin" <hpa@zytor.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Oleg Nesterov <oleg@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH 084/208] x86/fpu: Rename xsave.header::xstate_bv to 'xfeatures'
Date: Thu, 7 May 2015 14:22:09 +0200 [thread overview]
Message-ID: <20150507122209.GB15656@gmail.com> (raw)
In-Reply-To: <554A5D23.2030405@linux.intel.com>
* Dave Hansen <dave.hansen@linux.intel.com> wrote:
> On 05/05/2015 11:16 PM, Ingo Molnar wrote:
> > Btw., does Intel have any special plans with xstate compaction?
> >
> > AFAICS in Linux we just want to enable xfeat_mask_all to the max,
> > including compaction, and never really modify it (in the task's
> > lifetime).
>
> Special plans?
I.e. are there any plans beyond using it strictly for full state
save/restore.
> If we do an XRSTORS on it before we do an XSAVES, then we need to
> worry. But, if we do an XSAVES, the CPU will set it up for us.
>
> > I'm also wondering whether there will be any real 'holes' in the
> > xfeatures capability masks of future CPUs: right now xfeatures
> > tend to be already 'compacted' (because new CPUs tend to support
> > all xfeatures), so compaction mostly appears to be an academic
> > feature. Or is there already hardware out there where it matter?
>
> There is a hole in the SDM today. See section 2.6 in the currently
> released 054 version. I also know of actual hardware platforms with
> holes. *PLUS*, someone can always shot down CPUID bits in their
> hypervisor or with kernel command-line options.
I see, so MPX (bits 3 and 4) aren't there yet.
Btw., there's a new xfeature it appears:
XCR0.PKRU (bit 9): If 1, the XSAVE feature set can be used to manage
the PKRU register (see Section 2.7).
and bit 8 is a hole again.
Btw., regarding XCR0.PKRU: that enables 'Protection Keys' in the PTE
format. What's the main purpose of these keys? They seem to duplicate
the read/write bits in the PTE, with the exception that they don't
impact instruction fetches. So is this used to allow user-space to
execute but otherwise not read instructions?
Or some other purpose I missed?
In any case, these holes are really minor at the moment, and the
question is, what is the performance difference between a 'compactede'
XSAVE*/XRSTOR* pair, versus a standard format one?
> > Maybe once we get AVX512 in addition to MPX we can use compaction
> > materially: as there will be lots of tasks without MPX state but
> > with AVX512 state - in fact I suspect that will be the common
> > case.
>
> Right.
>
> But we'd need to get to a point where we are calling 'xsaves' with a
> Requested Feature BitMask (aka RFBM[]) that had holes in it. As it
> stands today, we always call it with RFBM=-1 and so we always have
> XCOMP_BV = XCR0.
XCOMP_BV must also have bit 63 set.
13.8.1
Standard Form of XRSTOR
The standard from of XRSTOR performs additional fault checking.
Either of the following conditions causes ageneral-protection
exception (#GP):
The XSTATE_BV field of the XSAVE header sets a bit that is not set
in XCR0. Bytes 23:8 of the XSAVE header are not all 0 (this implies
^^^^^^^^^^^^
that all bits in XCOMP_BV are 0).
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Note the part I underlined: all of XCOMP_BV has to be 0 for any
standard form of XRSTOR, and if we use a compacted form, bit 63 must
be set:
this is why bit 63 is a nonsensical interface: it being nonzero
already tells the hardware that we requested compaction ...
> We'd need to determine which fields are in the init state before we
> do an xsaves.
Why? I don't think that's necessary.
The way I read the SDM both the 'init' and the 'modified'
optimizations are mostly automatic: the CPU determines it
automatically when a state component is (or returned to!) init state,
and signals that via the relevant bit in XSTATE_BV being zeroed out.
This is what the SDM says about XSAVES (section 13.11 in the 054 SDM):
— If state component i is in its initial configuration, XSTATE_BV[i]
may be written with either 0 or 1.
so XSAVES itself performs the first step of the 'init optimization',
automatically: it will opportunistically write 0 to the relevant bit
in XSTATE_BV and won't save the state.
Once there's 0 in XSTATE_BV, put there by XSAVES, the XRSTOR
instruction is able to perform the other half of the optimization: by
not restoring it but initializing it (if needed).
XSAVES will also set up XSTATE_BV and XCOMP_BV so that XRSTOR does not
have to worry about it, it will do a compacted restore.
Thanks,
Ingo
next prev parent reply other threads:[~2015-05-07 12:22 UTC|newest]
Thread overview: 111+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-05-05 17:49 [PATCH 000/208] big x86 FPU code rewrite Ingo Molnar
2015-05-05 17:49 ` [PATCH 080/208] x86/fpu: Rename 'xstate_features' to 'xfeatures_nr' Ingo Molnar
2015-05-05 17:49 ` [PATCH 081/208] x86/fpu: Move XCR0 manipulation to the FPU code proper Ingo Molnar
2015-05-05 17:49 ` [PATCH 082/208] x86/fpu: Clean up regset functions Ingo Molnar
2015-05-05 17:49 ` [PATCH 083/208] x86/fpu: Rename 'xsave_hdr' to 'header' Ingo Molnar
2015-05-05 17:49 ` [PATCH 084/208] x86/fpu: Rename xsave.header::xstate_bv to 'xfeatures' Ingo Molnar
2015-05-05 17:57 ` Dave Hansen
2015-05-05 18:16 ` Ingo Molnar
2015-05-05 18:25 ` Dave Hansen
2015-05-06 6:16 ` Ingo Molnar
2015-05-06 12:46 ` Ingo Molnar
2015-05-06 15:09 ` Dave Hansen
2015-05-07 11:46 ` Ingo Molnar
2015-05-06 18:27 ` Dave Hansen
2015-05-07 10:59 ` Borislav Petkov
2015-05-07 12:22 ` Ingo Molnar [this message]
2015-05-07 14:58 ` Dave Hansen
2015-05-07 15:33 ` Ingo Molnar
2015-05-07 15:58 ` Dave Hansen
2015-05-07 19:35 ` Ingo Molnar
2015-05-05 17:49 ` [PATCH 085/208] x86/fpu: Clean up and fix MXCSR handling Ingo Molnar
2015-05-05 17:49 ` [PATCH 086/208] x86/fpu: Rename regset FPU register accessors Ingo Molnar
2015-05-05 17:49 ` [PATCH 087/208] x86/fpu: Explain the AVX register layout in the xsave area Ingo Molnar
2015-05-05 17:49 ` [PATCH 088/208] x86/fpu: Improve the __sanitize_i387_state() documentation Ingo Molnar
2015-05-05 17:49 ` [PATCH 089/208] x86/fpu: Rename fpu->has_fpu to fpu->fpregs_active Ingo Molnar
2015-05-05 17:49 ` [PATCH 090/208] x86/fpu: Rename __thread_set_has_fpu() to __fpregs_activate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 091/208] x86/fpu: Rename __thread_clear_has_fpu() to __fpregs_deactivate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 092/208] x86/fpu: Rename __thread_fpu_begin() to fpregs_activate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 093/208] x86/fpu: Rename __thread_fpu_end() to fpregs_deactivate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 094/208] x86/fpu: Remove fpstate_xstate_init_size() boot quirk Ingo Molnar
2015-05-05 17:49 ` [PATCH 095/208] x86/fpu: Remove xsave_init() bootmem allocations Ingo Molnar
2015-05-05 17:49 ` [PATCH 096/208] x86/fpu: Make setup_init_fpu_buf() run-once explicitly Ingo Molnar
2015-05-05 17:49 ` [PATCH 097/208] x86/fpu: Remove 'init_xstate_buf' bootmem allocation Ingo Molnar
2015-07-14 19:46 ` 4.2-rc2: early boot memory corruption from FPU rework Dave Hansen
2015-07-15 1:25 ` H. Peter Anvin
2015-07-15 11:07 ` Ingo Molnar
2015-07-16 0:34 ` [REGRESSION] " Dave Hansen
2015-07-16 2:39 ` Linus Torvalds
2015-07-16 2:51 ` Linus Torvalds
2015-07-17 7:45 ` Ingo Molnar
2015-07-17 8:51 ` Ingo Molnar
2015-07-17 16:58 ` Dave Hansen
2015-07-17 19:32 ` Ingo Molnar
2015-07-17 20:01 ` Dave Hansen
2015-05-05 17:49 ` [PATCH 098/208] x86/fpu: Split fpu__cpu_init() into early-boot and cpu-boot parts Ingo Molnar
2015-05-05 17:49 ` [PATCH 099/208] x86/fpu: Make the system/cpu init distinction clear in the xstate code as well Ingo Molnar
2015-05-05 17:49 ` [PATCH 100/208] x86/fpu: Move CPU capability check into fpu__init_cpu_xstate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 101/208] x86/fpu: Move legacy check to fpu__init_system_xstate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 102/208] x86/fpu: Propagate once per boot quirk into fpu__init_system_xstate() Ingo Molnar
2015-05-05 17:49 ` [PATCH 103/208] x86/fpu: Remove xsave_init() Ingo Molnar
2015-05-05 17:49 ` [PATCH 104/208] x86/fpu: Do fpu__init_system_xstate only from fpu__init_system() Ingo Molnar
2015-05-05 17:49 ` [PATCH 105/208] x86/fpu: Set up the legacy FPU init image " Ingo Molnar
2015-05-05 17:49 ` [PATCH 106/208] x86/fpu: Remove setup_init_fpu_buf() call from eager_fpu_init() Ingo Molnar
2015-05-05 17:49 ` [PATCH 107/208] x86/fpu: Move all eager-fpu setup code to eager_fpu_init() Ingo Molnar
2015-05-05 17:50 ` [PATCH 108/208] x86/fpu: Move eager_fpu_init() to fpu/init.c Ingo Molnar
2015-05-05 17:50 ` [PATCH 109/208] x86/fpu: Clean up eager_fpu_init() and rename it to fpu__ctx_switch_init() Ingo Molnar
2015-05-05 17:50 ` [PATCH 110/208] x86/fpu: Split fpu__ctx_switch_init() into _cpu() and _system() portions Ingo Molnar
2015-05-05 17:50 ` [PATCH 111/208] x86/fpu: Do CLTS fpu__init_system() Ingo Molnar
2015-05-05 17:50 ` [PATCH 112/208] x86/fpu: Move the fpstate_xstate_init_size() call into fpu__init_system() Ingo Molnar
2015-05-05 17:50 ` [PATCH 113/208] x86/fpu: Call fpu__init_cpu_ctx_switch() from fpu__init_cpu() Ingo Molnar
2015-05-05 17:50 ` [PATCH 114/208] x86/fpu: Do system-wide setup from fpu__detect() Ingo Molnar
2015-05-05 17:50 ` [PATCH 115/208] x86/fpu: Remove fpu__init_cpu_ctx_switch() call from fpu__init_system() Ingo Molnar
2015-05-05 17:50 ` [PATCH 116/208] x86/fpu: Simplify fpu__cpu_init() Ingo Molnar
2015-05-05 17:50 ` [PATCH 117/208] x86/fpu: Factor out fpu__init_cpu_generic() Ingo Molnar
2015-05-05 17:50 ` [PATCH 118/208] x86/fpu: Factor out fpu__init_system_generic() Ingo Molnar
2015-05-05 17:50 ` [PATCH 119/208] x86/fpu: Factor out fpu__init_system_early_generic() Ingo Molnar
2015-05-05 17:50 ` [PATCH 120/208] x86/fpu: Move !FPU check ingo fpu__init_system_early_generic() Ingo Molnar
2015-05-05 17:50 ` [PATCH 121/208] x86/fpu: Factor out FPU bug checks into fpu/bugs.c Ingo Molnar
2015-05-05 17:50 ` [PATCH 122/208] x86/fpu: Make check_fpu() init ordering independent Ingo Molnar
2015-05-05 17:50 ` [PATCH 123/208] x86/fpu: Move fpu__init_system_early_generic() out of fpu__detect() Ingo Molnar
2015-05-05 17:50 ` [PATCH 124/208] x86/fpu: Remove the extra fpu__detect() layer Ingo Molnar
2015-05-05 17:50 ` [PATCH 125/208] x86/fpu: Rename fpstate_xstate_init_size() to fpu__init_system_xstate_size_legacy() Ingo Molnar
2015-05-05 17:50 ` [PATCH 126/208] x86/fpu: Reorder init methods Ingo Molnar
2015-05-05 17:50 ` [PATCH 127/208] x86/fpu: Add more comments to the FPU init code Ingo Molnar
2015-05-05 17:50 ` [PATCH 128/208] x86/fpu: Move fpu__save() to fpu/internals.h Ingo Molnar
2015-05-05 17:50 ` [PATCH 129/208] x86/fpu: Uninline kernel_fpu_begin()/end() Ingo Molnar
2015-05-05 17:50 ` [PATCH 130/208] x86/fpu: Move various internal function prototypes to fpu/internal.h Ingo Molnar
2015-05-05 17:50 ` [PATCH 131/208] x86/fpu: Uninline the irq_ts_save()/restore() functions Ingo Molnar
2015-05-05 17:50 ` [PATCH 132/208] x86/fpu: Rename fpu_save_init() to copy_fpregs_to_fpstate() Ingo Molnar
2015-05-05 17:50 ` [PATCH 133/208] x86/fpu: Optimize copy_fpregs_to_fpstate() by removing the FNCLEX synchronization with FP exceptions Ingo Molnar
2015-05-05 17:50 ` [PATCH 134/208] x86/fpu: Simplify FPU handling by embedding the fpstate in task_struct (again) Ingo Molnar
2015-05-05 17:50 ` [PATCH 135/208] x86/fpu: Remove failure paths from fpstate-alloc low level functions Ingo Molnar
2015-05-05 17:50 ` [PATCH 136/208] x86/fpu: Remove failure return from fpstate_alloc_init() Ingo Molnar
2015-05-05 17:50 ` [PATCH 137/208] x86/fpu: Rename fpstate_alloc_init() to fpstate_init_curr() Ingo Molnar
2015-05-05 17:50 ` [PATCH 138/208] x86/fpu: Simplify fpu__unlazy_stopped() error handling Ingo Molnar
2015-05-05 17:50 ` [PATCH 139/208] x86/fpu, kvm: Simplify fx_init() Ingo Molnar
2015-05-05 17:50 ` [PATCH 140/208] x86/fpu: Simplify fpstate_init_curr() usage Ingo Molnar
2015-05-05 17:50 ` [PATCH 141/208] x86/fpu: Rename fpu__unlazy_stopped() to fpu__activate_stopped() Ingo Molnar
2015-05-05 17:50 ` [PATCH 142/208] x86/fpu: Factor out FPU hw activation/deactivation Ingo Molnar
2015-05-05 17:50 ` [PATCH 143/208] x86/fpu: Simplify __save_fpu() Ingo Molnar
2015-05-05 17:50 ` [PATCH 144/208] x86/fpu: Eliminate __save_fpu() Ingo Molnar
2015-05-05 17:50 ` [PATCH 145/208] x86/fpu: Simplify fpu__save() Ingo Molnar
2015-05-05 17:50 ` [PATCH 146/208] x86/fpu: Optimize fpu__save() Ingo Molnar
2015-05-05 17:50 ` [PATCH 147/208] x86/fpu: Optimize fpu_copy() Ingo Molnar
2015-05-05 17:50 ` [PATCH 148/208] x86/fpu: Optimize fpu_copy() some more on lazy switching systems Ingo Molnar
2015-05-05 17:50 ` [PATCH 149/208] x86/fpu: Rename fpu/xsave.h to fpu/xstate.h Ingo Molnar
2015-05-05 17:50 ` [PATCH 150/208] x86/fpu: Rename fpu/xsave.c to fpu/xstate.c Ingo Molnar
2015-05-05 17:50 ` [PATCH 151/208] x86/fpu: Introduce cpu_has_xfeatures(xfeatures_mask, feature_name) Ingo Molnar
2015-05-05 22:15 ` Yu, Fenghua
2015-05-06 5:00 ` Ingo Molnar
2015-05-05 17:50 ` [PATCH 152/208] x86/fpu: Simplify print_xstate_features() Ingo Molnar
2015-05-05 17:50 ` [PATCH 153/208] x86/fpu: Enumerate xfeature bits Ingo Molnar
2015-05-05 17:50 ` [PATCH 154/208] x86/fpu: Move xfeature type enumeration to fpu/types.h Ingo Molnar
2015-05-05 17:50 ` [PATCH 155/208] x86/fpu, crypto x86/camellia_aesni_avx: Simplify the camellia_aesni_init() xfeature checks Ingo Molnar
2015-05-05 17:50 ` [PATCH 156/208] x86/fpu, crypto x86/sha256_ssse3: Simplify the sha256_ssse3_mod_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 157/208] x86/fpu, crypto x86/camellia_aesni_avx2: Simplify the camellia_aesni_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 158/208] x86/fpu, crypto x86/twofish_avx: Simplify the twofish_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 159/208] x86/fpu, crypto x86/serpent_avx: Simplify the serpent_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 160/208] x86/fpu, crypto x86/cast5_avx: Simplify the cast5_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 161/208] x86/fpu, crypto x86/sha512_ssse3: Simplify the sha512_ssse3_mod_init() " Ingo Molnar
2015-05-05 17:50 ` [PATCH 162/208] x86/fpu, crypto x86/cast6_avx: Simplify the cast6_init() " Ingo Molnar
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=20150507122209.GB15656@gmail.com \
--to=mingo@kernel.org \
--cc=bp@alien8.de \
--cc=dave.hansen@linux.intel.com \
--cc=fenghua.yu@intel.com \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@amacapital.net \
--cc=mingo2.kernel.org@gmail.com \
--cc=oleg@redhat.com \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.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).