[Qemu-devel] Character order of ATA identity

[Qemu-devel] Character order of ATA identity

[乔天香]

[-- Attachment #1: Type: text/plain, Size: 1086 bytes --]

Hi, all

I'm learning operating system and trying to write a little toy kernel.
Before doing disk I/O, I issued the *ATA identify* command to obtain
information about the master device on the primary channel. When I ran my
code with QEMU, I found it gives some ASCII string fields in a wrong byte
order, for example, the model name is given as:

    EQUMH RADDSI K

but actually it should be

    QEMU HARDDISK

So I referred to the source code file `hw/ide/core.c` and found a function
doing this:


  61 static void padstr(char *str, const char *src, int len)
  62 {
  63     int i, v;
  64     for(i = 0; i < len; i++) {
  65         if (*src)
  66             v = *src++;
  67         else
  68             v = ' ';
  69         str[i^1] = v;
  70     }
  71 }


And it is called like:

     112     padstr((char *)(p + 27), s->drive_model_str, 40); /* model */

Now I'm wondering why it does this "byte swapping"? I read the ATA
specification about this *ATA identify* command and didn't find anything
related to byte order. Is it required by hardware?

Thanks for your kind help.

[-- Attachment #2: Type: text/html, Size: 2188 bytes --]

Re: [Qemu-devel] Character order of ATA identity

[John Snow]

On 04/16/2015 02:07 PM, 乔天香 wrote:
> Hi, all
>

Hi!

> I'm learning operating system and trying to write a little toy kernel.
> Before doing disk I/O, I issued the *ATA identify* command to obtain
> information about the master device on the primary channel. When I ran
> my code with QEMU, I found it gives some ASCII string fields in a wrong
> byte order, for example, the model name is given as:
>
> EQUMH RADDSI K
>
> but actually it should be
>
>      QEMU HARDDISK
>
> So I referred to the source code file `hw/ide/core.c` and found a
> function doing this:
>
>
>    61 static void padstr(char *str, const char *src, int len)
>    62 {
>    63     int i, v;
>    64     for(i = 0; i < len; i++) {
>    65         if (*src)
>    66             v = *src++;
>    67         else
>    68             v = ' ';
>    69         str[i^1] = v;
>    70     }
>    71 }
>
>
> And it is called like:
>
>       112     padstr((char *)(p + 27), s->drive_model_str, 40); /* model */
>
> Now I'm wondering why it does this "byte swapping"? I read the ATA
> specification about this *ATA identify* command and didn't find anything
> related to byte order. Is it required by hardware?
>
> Thanks for your kind help.

The reason is that the ASCII data is, for whatever reason, stored as an 
array of 2 byte chunks. Any historical reasons for doing so are not 
known to me, but:

Please check out ATA8 ATA/ATAPI Command Set 3, Revision 1b (or whichever 
ATA spec you have handy)

Section 3.3.10 "ATA string convention"
'Each pair of bytes in an ATA string is swapped as shown in table 5.'

Why?

It's a mystery to me.

Please don't hesitate to comment on any other oddities you come across 
within the IDE system.

Thanks,
--js