Re: [PATCH 1/2] cpqfc: fix for "Using too much stach" in 2.6 kernel

[Rolf Eike Beer <eike-kernel@sf-tec.de>]

Am Donnerstag, 4. August 2005 17:40 schrieb Dave Jones:
>On Thu, Aug 04, 2005 at 11:38:30AM +0200, Rolf Eike Beer wrote:
> > >+	  ulFibreFrame = kmalloc((2048/4), GFP_KERNEL);
> >
> > The size bug was already found by Dave Jones. This never should be
> > written this way (not your fault). The array should have been
> > [2048/sizeof(ULONG)].
>
>wasteful. We only ever use 2048 bytes of this array, so doubling
>its size on 64bit is pointless, unless you make changes later on
>in the driver. (Which I think don't make sense, as we just copy
>32 64byte chunks).

No, this is how it should have been before. This way it would have been clear 
where the magic 4 came from.

>Ermm, actually this looks totally bogus..
>CpqTsGetSFQEntry() ...
>
>    if( total_bytes <= 2048 )
>    {
>      memcpy( ulDestPtr,
>              &fcChip->SFQ->QEntry[consumerIndex],
>              64 );  // each SFQ entry is 64 bytes
>      ulDestPtr += 16;   // advance pointer to next 64 byte block
>    }
>
>we're trashing the last 48 bytes of every copy we make.
>Does this driver even work ?

No, ulDestPtr ist ULONG* so we increase it by sizeof(ULONG)*16 which is 64. 
This is one of the places I was talking about where people might miss what's 
going on. ;) IMHO it makes absolutely no sense to use a ULONG* at this place.

Eike

Attachment: pgp5KMktgVrY6.pgp
Description: PGP signature