xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: Andrew Cooper <andrew.cooper3@citrix.com>
To: Aravind Gopalakrishnan <aravind.gopalakrishnan@amd.com>,
	JBeulich@suse.com, xen-devel@lists.xen.org
Cc: boris.ostrovsky@oracle.com, keir@xen.org
Subject: Re: [PATCH V2] x86, amd_ucode: Safeguard against #GP
Date: Wed, 28 May 2014 00:47:17 +0100	[thread overview]
Message-ID: <53852405.9010704@citrix.com> (raw)
In-Reply-To: <1401215048-17154-1-git-send-email-aravind.gopalakrishnan@amd.com>

On 27/05/2014 19:24, Aravind Gopalakrishnan wrote:
> When HW tries to load a corrupted patch, it generates #GP
> and hangs the system. Use wrmsr_safe instead so that we
> fail to load microcode gracefully.
>
> Also, massage error handling around apply_microcode to keep
> in tune with error handling style of other parts of the code.
>
> Example on a Fam15h system-
> (XEN) microcode: CPU0 collect_cpu_info: patch_id=0x6000626
> (XEN) microcode: CPU0 size 7870, block size 2586 offset 76 equivID
> 0x6012 rev 0x6000637
> (XEN) microcode: CPU0 found a matching microcode update with version
> 0x6000637 (current=0x6000626)
> (XEN) traps.c:3073: GPF (0000): ffff82d08016f682 -> ffff82d08022d9f8
> (XEN) microcode: CPU0 update from revision 0x6000637 to 0x6000626 failed
>                                            ^^^^^^^^^^^^^^^^^^^^^^
> As shown, the log message above has the two revisions reversed. Fix this
>
> Changes in V2:
>     - Do not ignore return value from wrmsr_safe
>     - Flip revision numbers as shown above
>
> Signed-off-by: Aravind Gopalakrishnan <aravind.gopalakrishnan@amd.com>
> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>

I thought we had identified that the hangs were to do with your use of
'noreboot' on the Xen command line.

~Andrew


> ---
>  xen/arch/x86/microcode_amd.c |   25 +++++++++++++++++++------
>  1 file changed, 19 insertions(+), 6 deletions(-)
>
> diff --git a/xen/arch/x86/microcode_amd.c b/xen/arch/x86/microcode_amd.c
> index e83f4b6..1db8a0d 100644
> --- a/xen/arch/x86/microcode_amd.c
> +++ b/xen/arch/x86/microcode_amd.c
> @@ -178,32 +178,39 @@ static int apply_microcode(int cpu)
>      uint32_t rev;
>      struct microcode_amd *mc_amd = uci->mc.mc_amd;
>      struct microcode_header_amd *hdr;
> +    int error = -EINVAL;
>  
>      /* We should bind the task to the CPU */
>      BUG_ON(raw_smp_processor_id() != cpu);
>  
>      if ( mc_amd == NULL )
> -        return -EINVAL;
> +        goto apply_err1;
>  
>      hdr = mc_amd->mpb;
>      if ( hdr == NULL )
> -        return -EINVAL;
> +        goto apply_err1;
>  
>      spin_lock_irqsave(&microcode_update_lock, flags);
>  
> -    wrmsrl(MSR_AMD_PATCHLOADER, (unsigned long)hdr);
> +    error = wrmsr_safe(MSR_AMD_PATCHLOADER, (unsigned long)hdr);
>  
>      /* get patch id after patching */
>      rdmsrl(MSR_AMD_PATCHLEVEL, rev);
>  
>      spin_unlock_irqrestore(&microcode_update_lock, flags);
>  
> +    /* Catch HW patch application failure */
> +    if ( error ) {
> +        printk(KERN_ERR "microcode: CPU%d ucode patch application failed HW tests. "
> +               "HW returned #GP\n", cpu);
> +        goto apply_err2;

This...

> +    }
> +
>      /* check current patch id and patch's id for match */
>      if ( rev != hdr->patch_id )
>      {
> -        printk(KERN_ERR "microcode: CPU%d update from revision "
> -               "%#x to %#x failed\n", cpu, hdr->patch_id, rev);
> -        return -EIO;
> +        error = -EIO;
> +        goto apply_err2;
>      }
>  
>      printk(KERN_WARNING "microcode: CPU%d updated from revision %#x to %#x\n",
> @@ -212,6 +219,12 @@ static int apply_microcode(int cpu)
>      uci->cpu_sig.rev = rev;
>  
>      return 0;
> +
> +apply_err2:
> +    printk(KERN_ERR "microcode: CPU%d update from revision "
> +           "%#x to %#x failed\n", cpu, rev, hdr->patch_id);

... combined with this will result in two error messages being printed.

This seems over overkill for the circumstance.

~Andrew.

> +apply_err1:
> +    return error;
>  }
>  
>  static int get_ucode_from_buffer_amd(

  reply	other threads:[~2014-05-27 23:47 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-05-27 18:24 [PATCH V2] x86, amd_ucode: Safeguard against #GP Aravind Gopalakrishnan
2014-05-27 23:47 ` Andrew Cooper [this message]
2014-05-28 15:16   ` Aravind Gopalakrishnan
2014-05-28 17:56     ` boris ostrovsky
2014-05-30 16:01       ` Aravind Gopalakrishnan
2014-05-30 16:21         ` Andrew Cooper
2014-05-30 16:46           ` Aravind Gopalakrishnan
2014-06-02  7:31           ` Jan Beulich
2014-06-02 14:13             ` Boris Ostrovsky
2014-05-28  7:22 ` Jan Beulich

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=53852405.9010704@citrix.com \
    --to=andrew.cooper3@citrix.com \
    --cc=JBeulich@suse.com \
    --cc=aravind.gopalakrishnan@amd.com \
    --cc=boris.ostrovsky@oracle.com \
    --cc=keir@xen.org \
    --cc=xen-devel@lists.xen.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).