2.4.0 tulip bug (was: And oh, btw..)

2.4.0 tulip bug (was: And oh, btw..)

[Mikael Pettersson]

On Thu, 4 Jan 2001, Linus Torvalds wrote:

>Changes since the prerelease:
>...
>Matti Aarnio:
> - teach tulip driver about media types 5 and 6

This part of the patch introduces a bug in 2.4.0, as noticed by gcc:

media.c: In function `tulip_select_media':
media.c:268: warning: unused variable `csr15val'
media.c:268: warning: unused variable `csr15dir'
media.c:268: warning: unused variable `csr14val'
media.c:268: warning: unused variable `csr13val'
media.c:151: warning: `new_csr6' might be used uninitialized in this function

The last warning indicates a real problem. The patch adds a new
control flow path in which new_csr6 is _not_ assigned a value,
which causes the procedure's second last statement

	tp->csr6 = new_csr6 | (tp->csr6 & 0xfdff) | (tp->full_duplex ? 0x0200 : 0);

to 'or' random bits into tp->csr6.

The patch also adds four unused variables, which looks rather fishy.

/Mikael
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: 2.4.0 tulip bug (was: And oh, btw..)

[Matti Aarnio]

On Fri, Jan 05, 2001 at 04:23:13PM +0100, Mikael Pettersson wrote:
> >Changes since the prerelease:
> >...
> >Matti Aarnio:
> > - teach tulip driver about media types 5 and 6
>
> This part of the patch introduces a bug in 2.4.0, as noticed by gcc:

	Oddly enough the beast works with my test specimen,
	but real analysis is still incomplete by Donald et.al.

	So far we think that if the card has MII (type 1) media,
	it could safely ignore types 5 and 6.

> The patch also adds four unused variables, which looks rather fishy.

	Indeed, it is copy&paste from somewhere else, which may
	indeed have different control flows.

	However unless you have esoteric (taiwanese) special hardware
	with those media types, it won't affect you.

> /Mikael

/Matti Aarnio
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/