Re: SATA status report updated

[Jeff Garzik <jgarzik@pobox.com>]

Simon Oosthoek wrote:
> I know Mandriva is on the ball and a bug with some information and an 
> updated patch is on the kernel bugme...
> 
> http://qa.mandriva.com/show_bug.cgi?id=17654
> http://bugme.osdl.org/show_bug.cgi?id=4192
> 
> I'd say it's important to get some proper fix in a distribution soon (so 
> I can use my new PC ;-)


That's not an updated patch.  That's the patch that duplicates kernel 
infrastructure, implementing things in the driver that should instead be 
implemented in libata core.

That's how Windows drivers are written: work around the OS, rather than 
fix it.

Here is a list of problems with the patch.  I'll paste this into the bug 
as well:

1) duplicates SATA phy reset

2) abuses infrastructure to support PATA, rather than doing it properly. 
  doing it properly involves an approach similar to that found in the 
'promise-sata-pata' branch of libata-dev.git.  Same problem as Promise 
SATA+PATA, with the same solution.

3) duplicates ATA bus reset, except, does it poorly

4) duplicates ata_busy_sleep()

5) appears to do strange things with PATA devices, when one uses the 
->scr_write() and ->scr_read() hooks -- hooks used to talk to SATA PHYs 
(never PATA devices).

6) [maybe] sets DMA/PIO timings even for SATA devices.  This -may- be 
needed, depending on PATA<->SATA bridge presence in the host controller

7) Pads DMA to 32-bit boundary.  Should be done in libata core, this is 
needed for all host controllers.

8) The DMA pad code is very buggy.  It uses the dma_map_single() to map 
a buffer, but never synchronizes nor flushes the buffer.  This can and 
will lead to data corruption, particularly on x86-64 platform.

Regards,

	Jeff