why does UATA/133 == UATA/100 on amd64?

[Alexander Motin]

fbsdmail at dnswatch.com wrote:
> On Fri, June 4, 2010 10:58 pm, Alexander Motin wrote:
>> Peter Jeremy wrote:
>>> On 2010-Jun-04 16:36:08 -0700, fbsdmail at dnswatch.com wrote:
>>>
>>>> After _finally_ making the correct decisions to install amd64 on an
>>>> AMD64 system. I was able to make/build/install world && kernel, I see
>>>> a difference in drive recognition.
>>> Can you please do a verbose boot and post the resultant dmesg somewhere
>>>  (preferably with your USB DVD drive connected).
>>>
>>>> kernel: ata3-master: pio=PIO4 wdma=WDMA2 udma=UDMA133 cable=40 wire
>>>> kernel: ad6: 476940MB <Seagate ST3500630AS 3.AAK> at ata3-master
>>>> SATA300
>>>>
>>>> kernel: ata3-master: pio=PIO4 wdma=WDMA2 udma=UDMA133 cable=40 wire
>>>> kernel: ad6: setting UDMA100
>>>> kernel: ad6: 476940MB <Seagate ST3500630AS 3.AAK> at ata3-master
>>>> UDMA100
>>>> SATA 3Gb/s
>>>>
>>> The 'UDMA' numbers are meaningless for SATA controllers/drives.
>>>
>> The 'UDMA' numbers are meaningless for _native_ SATA controllers/drives.
>>
>> They may be not meaningless for legacy SATA devices, using SATA->PATA
>> bridge inside. Some bridges do not support UDMA133 on PATA part, so ata(4)
>> prefers not to use it. But in this case it is indeed meaningless.
> 
> If it's not already apparent. The board has an AMD 880G chipset, that
> provides RAID support on 6 ports @ 6GBs. Now, from a purely logistical
> standpoint. The numbers _can't_ be meaningless. It's clear that the kernel
> is making a "judgment call" here: kernel: ad6: setting UDMA100

It is impossible to detect SATA->PATA bridge presence, so kernel has to
always follow worst scenario. But as I have said, for this particular
device this value affects nothing.

> The "judgment call" on both GENERIC/i386, and GENERIC/amd64 was never
> made. The capability of both the port && the drive were accepted. Both
> cases were booted using "verbose" (5). Please understand, I'm not
> attempting to be argumentative here. I just observe this to be true.
> In other words; it must have _some_ meaning - no?

I have feeling that you have updated your sources while building custom
kernel. I can't explain difference you have shown by other reasons.

-- 
Alexander Motin