Re: [PATCH] ide_dma_speed() fixes

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Hello.

Andrew Morton wrote:

>>    ide_dma_speed() fails to actually honor the IDE drivers' mode support 
>> masks) because of the bogus checks -- thus, selecting the DMA transfer mode 
>> that the driver explicitly refuses to support is possible. Additionally, there 
>> is no check for validity of the UltraDMA mode data in the drive ID, and the 
>> function is misdocumented.

> drivers/ide/ide-lib.c: In function `ide_dma_speed':
> drivers/ide/ide-lib.c:86: warning: `ultra_mask' might be used uninitialized in this function

> Looks like a real bug to me - it depends up on the values of `mode' and
> id->field_value.

> Anyway, I'll drop it, please review and fix.  I assume that warning was
> occurring for you as well - please spend more time over these things. 
> Especially when working on IDE, where bugs are slow to show themselves and
> have particularly bad consequences.

    That's what gcc thinks. The code is 100% correct -- it will never reach 
the switch statement with mode > 0 (in which case ultra_mask isn't used) and 
ultra_mask unitialized. I may add an explicit initializer in the declaration 
if you like...

MBR, Sergei