Re: [PATCH 11/11] x86/ucode/amd: Rework parsing logic in cpu_request_microcode()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 31.03.2020 17:19, Andrew Cooper wrote:
> On 31/03/2020 16:07, Jan Beulich wrote:
>> On 31.03.2020 12:05, Andrew Cooper wrote:
>>> @@ -269,55 +265,25 @@ static int apply_microcode(const struct 
>>> microcode_patch *patch)
>>>      return 0;
>>>  }
>>>  
>>> -static int scan_equiv_cpu_table(
>>> -    const void *data,
>>> -    size_t size_left,
>>> -    size_t *offset)
>>> +static int scan_equiv_cpu_table(const struct container_equiv_table *et)
>>>  {
>>>      const struct cpu_signature *sig = &this_cpu(cpu_sig);
>>> -    const struct mpbhdr *mpbuf;
>>> -    const struct equiv_cpu_entry *eq;
>>> -    unsigned int i, nr;
>>> -
>>> -    if ( size_left < (sizeof(*mpbuf) + 4) ||
>>> -         (mpbuf = data + *offset + 4,
>>> -          size_left - sizeof(*mpbuf) - 4 < mpbuf->len) )
>>> -    {
>>> -        printk(XENLOG_WARNING "microcode: No space for equivalent cpu 
>>> table\n");
>>> -        return -EINVAL;
>>> -    }
>>> -
>>> -    *offset += mpbuf->len + CONT_HDR_SIZE; /* add header length */
>>> -
>>> -    if ( mpbuf->type != UCODE_EQUIV_CPU_TABLE_TYPE )
>>> -    {
>>> -        printk(KERN_ERR "microcode: Wrong microcode equivalent cpu table 
>>> type field\n");
>>> -        return -EINVAL;
>>> -    }
>>> -
>>> -    if ( mpbuf->len == 0 || mpbuf->len % sizeof(*eq) ||
>>> -         (eq = (const void *)mpbuf->data,
>>> -          nr = mpbuf->len / sizeof(*eq),
>>> -          eq[nr - 1].installed_cpu) )
>> Did this last check get lost? I can't seem to be able to identify
>> any possible replacement.
> 
> Given the lack of a spec, I'm unsure whether to keep it or not.
> 
> It is necessary in the backport of patch 1, because find_equiv_cpu_id()
> doesn't have mpbuf->len to hand, and relies on the sentinel to find the
> end of the table.
> 
> OTOH, the new logic will cope perfectly well without a sentinel.

Okay.

>>>  static struct microcode_patch *cpu_request_microcode(const void *buf, 
>>> size_t size)
>>>  {
>>>      const struct microcode_patch *saved = NULL;
>>>      struct microcode_patch *patch = NULL;
>>> -    size_t offset = 0, saved_size = 0;
>>> +    size_t saved_size = 0;
>>>      int error = 0;
>>> -    unsigned int cpu = smp_processor_id();
>>> -    const struct cpu_signature *sig = &per_cpu(cpu_sig, cpu);
>>>  
>>> -    if ( size < 4 ||
>>> -         *(const uint32_t *)buf != UCODE_MAGIC )
>>> +    while ( size )
>>>      {
>>> -        printk(KERN_ERR "microcode: Wrong microcode patch file magic\n");
>>> -        error = -EINVAL;
>>> -        goto out;
>>> -    }
>>> -
>>> -    /*
>>> -     * Multiple container file support:
>>> -     * 1. check if this container file has equiv_cpu_id match
>>> -     * 2. If not, fast-fwd to next container file
>>> -     */
>>> -    while ( offset < size )
>>> -    {
>>> -        error = scan_equiv_cpu_table(buf, size - offset, &offset);
>>> -
>>> -        if ( !error || error != -ESRCH )
>>> -            break;
>>> +        const struct container_equiv_table *et;
>>> +        bool skip_ucode;
>>>  
>>> -        error = container_fast_forward(buf, size - offset, &offset);
>>> -        if ( error == -ENODATA )
>>> +        if ( size < 4 || *(const uint32_t *)buf != UCODE_MAGIC )
>>>          {
>>> -            ASSERT(offset == size);
>>> +            printk(XENLOG_ERR "microcode: Wrong microcode patch file 
>>> magic\n");
>>> +            error = -EINVAL;
>>>              break;
>>>          }
>>> -        if ( error )
>>> +
>>> +        /* Move over UCODE_MAGIC. */
>>> +        buf  += 4;
>>> +        size -= 4;
>>> +
>>> +        if ( size < sizeof(*et) ||
>>> +             (et = buf)->type != UCODE_EQUIV_CPU_TABLE_TYPE ||
>>> +             size - sizeof(*et) < et->len ||
>>> +             et->len % sizeof(et->eq[0]) )
>>>          {
>>> -            printk(KERN_ERR "microcode: CPU%d incorrect or corrupt 
>>> container file\n"
>>> -                   "microcode: Failed to update patch level. "
>>> -                   "Current lvl:%#x\n", cpu, sig->rev);
>>> +            printk(XENLOG_ERR "microcode: Bad equivalent cpu table\n");
>>> +            error = -EINVAL;
>>>              break;
>>>          }
>>> -    }
>>>  
>>> -    if ( error )
>>> -    {
>>> -        /*
>>> -         * -ENODATA here means that the blob was parsed fine but no 
>>> matching
>>> -         * ucode was found. Don't return it to the caller.
>>> -         */
>>> -        if ( error == -ENODATA )
>>> -            error = 0;
>>> -
>>> -        goto out;
>>> -    }
>>> +        /* Move over the Equiv table. */
>>> +        buf  += sizeof(*et) + et->len;
>>> +        size -= sizeof(*et) + et->len;
>>> +
>>> +        error = scan_equiv_cpu_table(et);
>>> +        if ( error && error != -ESRCH )
>>> +            break;
>> With this the only non-zero value left for error is -ESRCH.
>> Hence ...
>>
>>> +        /* -ESRCH means no applicable microcode in this container. */
>>> +        skip_ucode = error == -ESRCH;
>> ... perhaps omit the "== -ESRCH" here, moving the comment up
>> ahead of the if()?
> 
> That doesn't work, because you've got to reset error to 0 somewhere (to
> avoid it leaking out if you don't find suitable microcode), and it can't
> be before checking for errors in general.  It can't easily become a
> conditional because skip_ucode needs setting unconditionally.

I don't follow - what's wrong with

        /* -ESRCH means no applicable microcode in this container. */
        if ( error && error != -ESRCH )
           break;
        skip_ucode = error;
        error = 0;

?

> I have been debating quite heavily whether -ESRCH is best here, or using
> -ve, 0 and 1.  However, this doesn't lead to prettier code AFAICT, and
> gains an ambiguous use for a variable named "error".

I'm fine with that choice of yours.

Jan