broken 0 sized memset checks?

broken 0 sized memset checks?

[Dave Jones]

I ran sparse over the current linux kernel, and noticed this..

drivers/net/chelsio/sge.c:341:9: warning: memset with byte count of 0
drivers/net/chelsio/sge.c:469:9: warning: memset with byte count of 0

The first..

        q->size = p->freelQ_size[i];
        ....
        size = sizeof(struct freelQ_ce) * q->size;
        q->centries = kmalloc(size, GFP_KERNEL);
        if (!q->centries)
            goto err_no_mem;
        memset(q->centries, 0, size);

Looks valid enough, as does the second..

        q->size = p->cmdQ_size[i];
        ....
        size = sizeof(struct cmdQ_ce) * q->size;
        q->centries = kmalloc(size, GFP_KERNEL);
        if (!q->centries)
            goto err_no_mem;
        memset(q->centries, 0, size);

Sparse bug ?

	Dave

Re: broken 0 sized memset checks?

[Linus Torvalds]

On Tue, 5 Sep 2006, Dave Jones wrote:
>
> I ran sparse over the current linux kernel, and noticed this..
> 
> drivers/net/chelsio/sge.c:341:9: warning: memset with byte count of 0
> drivers/net/chelsio/sge.c:469:9: warning: memset with byte count of 0
> 
> The first..
> 
>         q->size = p->freelQ_size[i];
>         ....
>         size = sizeof(struct freelQ_ce) * q->size;
>         q->centries = kmalloc(size, GFP_KERNEL);
>         if (!q->centries)
>             goto err_no_mem;
>         memset(q->centries, 0, size);
>
> Sparse bug ?

I _suspect_ sparse has had some type-parsing bug, and thinks that 
"sizeof(struct freelQ_ce)" is zero for some reason, and then does all the 
constant folding (yeah, sparse is clever when it isn't really stupid ;), 
and that's what is going on.

I'll see if I can figure out what's up.

		Linus