QEMU-Devel Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Magnus Kulke <magnuskulke@linux.microsoft.com>
To: "Doru Blânzeanu" <dblanzeanu@linux.microsoft.com>
Cc: qemu-devel@nongnu.org, "Magnus Kulke" <magnuskulke@microsoft.com>,
	"Wei Liu" <wei.liu@kernel.org>, "Wei Liu" <liuwe@microsoft.com>,
	"Doru Blânzeanu" <dblanzeanu@microsoft.com>
Subject: Re: [PATCH v2 2/2] accel/mshv: enable amx tiles support for guests
Date: Thu, 15 Oct 2026 13:40:18 +0200	[thread overview]
Message-ID: <atC7onhsNjrADI1S@example.com> (raw)
In-Reply-To: <20260922142156.89306-3-dblanzeanu@linux.microsoft.com>

On Tue, Sep 22, 2026 at 05:21:56PM +0300, Doru Blânzeanu wrote:
> The AMX tiles feature uses a larger than 4kB area in xsave.
> This fix removes the code that disables AMX tiles in the partition and it
> adjusts the supported CPUID to advertise AMX tiles.
> 
> Signed-off-by: Doru Blânzeanu <dblanzeanu@linux.microsoft.com>
> ---
>  accel/mshv/mshv-all.c       | 18 +++++-------------
>  target/i386/mshv/mshv-cpu.c | 16 ----------------
>  2 files changed, 5 insertions(+), 29 deletions(-)
> 
> diff --git a/accel/mshv/mshv-all.c b/accel/mshv/mshv-all.c
> index 725ccd0511..99b2b3362a 100644
> --- a/accel/mshv/mshv-all.c
> +++ b/accel/mshv/mshv-all.c
> @@ -187,7 +187,6 @@ static int create_partition(int mshv_fd, int *vm_fd)
>  {
>      int ret;
>      uint64_t pt_flags, host_proc_features;
> -    union hv_partition_processor_xsave_features disabled_xsave_features;
>      union hv_partition_processor_features disabled_partition_features = {0};
>  
>      struct mshv_create_partition_v2 args = {0};
> @@ -200,17 +199,6 @@ static int create_partition(int mshv_fd, int *vm_fd)
>                 (1ULL << MSHV_PT_BIT_GPA_SUPER_PAGES) |
>                 (1ULL << MSHV_PT_BIT_CPU_AND_XSAVE_FEATURES);
>  
> -    /* enable all */
> -    disabled_xsave_features.as_uint64 = 0;
> -    /*
> -     * AMX TILE XSAVE state (XTILE_DATA) is 8KB, which exceeds the
> -     * current fixed 4KB XSAVE buffer size.
> -     */
> -    disabled_xsave_features.amx_tile_support = 1;
> -    disabled_xsave_features.amx_bf16_support = 1;
> -    disabled_xsave_features.amx_int8_support = 1;
> -    disabled_xsave_features.amx_fp16_support = 1;
> -
>      /*
>       * query host for supported processor features and disable unsupported
>       * features: (0 means supported, 1 means disabled, hence the negation)
> @@ -238,10 +226,14 @@ static int create_partition(int mshv_fd, int *vm_fd)
>      args.pt_cpu_fbanks[0] |= disabled_partition_features.as_uint64[0];
>      args.pt_cpu_fbanks[1] |= disabled_partition_features.as_uint64[1];
>  
> +    /*
> +     * for now, all the xsave features are enabled, the filtering is only done
> +     * at the CPUID level
> +     * TODO: filter the xsave features based on the CPUID derived from Cpu model
> +     */
>      /* populate args structure */
>      args.pt_flags = pt_flags;
>      args.pt_isolation = MSHV_PT_ISOLATION_NONE;
> -    args.pt_disabled_xsave = disabled_xsave_features.as_uint64;
>      args.pt_num_cpu_fbanks = MSHV_NUM_CPU_FEATURES_BANKS;
>  
>      ret = ioctl(mshv_fd, MSHV_CREATE_PARTITION, &args);
> diff --git a/target/i386/mshv/mshv-cpu.c b/target/i386/mshv/mshv-cpu.c
> index 5b845b3e4e..21a3534578 100644
> --- a/target/i386/mshv/mshv-cpu.c
> +++ b/target/i386/mshv/mshv-cpu.c
> @@ -2234,22 +2234,6 @@ uint32_t mshv_get_supported_cpuid(uint32_t func, uint32_t idx, int reg)
>           */
>          ret &= ~CPUID_7_0_ECX_LA57;
>      }
> -    if (func == 0x07 && idx == 0 && reg == R_EDX) {
> -        /*
> -         * AMX TILE XSAVE state (XTILE_DATA) is 8KB, which exceeds the
> -         * current fixed 4KB XSAVE buffer size. Filter until buffer
> -         * sizing is computed dynamically from CPUID.
> -         */
> -        ret &= ~CPUID_7_0_EDX_AMX_TILE;
> -        ret &= ~CPUID_7_0_EDX_AMX_BF16;
> -        ret &= ~CPUID_7_0_EDX_AMX_INT8;
> -    }
> -    if (func == 0x07 && idx == 1 && reg == R_EAX) {
> -        ret &= ~CPUID_7_1_EAX_AMX_FP16;
> -    }
> -    if (func == 0x07 && idx == 1 && reg == R_EDX) {
> -        ret &= ~CPUID_7_1_EDX_AMX_COMPLEX;
> -    }
>  
>      return ret;
>  }
> -- 
> 2.53.0

Reviewed-by: Magnus Kulke <magnuskulke@linux.microsoft.com>


      reply	other threads:[~2026-10-15 11:40 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-22 14:21 [PATCH v2 0/2] acce/mshv: size XSAVE buffers from the hypervisor Doru Blânzeanu
2026-09-22 14:21 ` [PATCH v2 1/2] accel/mshv: " Doru Blânzeanu
2026-10-15 11:39   ` Magnus Kulke
2026-09-22 14:21 ` [PATCH v2 2/2] accel/mshv: enable amx tiles support for guests Doru Blânzeanu
2026-10-15 11:40   ` Magnus Kulke [this message]

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=atC7onhsNjrADI1S@example.com \
    --to=magnuskulke@linux.microsoft.com \
    --cc=dblanzeanu@linux.microsoft.com \
    --cc=dblanzeanu@microsoft.com \
    --cc=liuwe@microsoft.com \
    --cc=magnuskulke@microsoft.com \
    --cc=qemu-devel@nongnu.org \
    --cc=wei.liu@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