linux-kernel - Re: [PATCH] of: address: Unify resource bounds overflow checking

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <87plp7r8ye.fsf@mail.lhotse>
Date: Fri, 13 Sep 2024 23:15:21 +1000
From: Michael Ellerman <mpe@...erman.id.au>
To: Thomas Weißschuh <thomas.weissschuh@...utronix.de>, Rob
 Herring
 <robh@...nel.org>, Saravana Kannan <saravanak@...gle.com>
Cc: devicetree@...r.kernel.org, linux-kernel@...r.kernel.org, Thomas
 Weißschuh <thomas.weissschuh@...utronix.de>,
 <linuxppc-dev@...ts.ozlabs.org> 
Subject: Re: [PATCH] of: address: Unify resource bounds overflow checking

Thomas Weißschuh <thomas.weissschuh@...utronix.de> writes:
> The members "start" and "end" of struct resource are of type
> "resource_size_t" which can be 32bit wide.
> Values read from OF however are always 64bit wide.
>
> Refactor the diff overflow checks into a helper function.
> Also extend the checks to validate each calculation step.
>
> Signed-off-by: Thomas Weißschuh <thomas.weissschuh@...utronix.de>
> ---
>  drivers/of/address.c | 45 ++++++++++++++++++++++++++-------------------
>  1 file changed, 26 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/of/address.c b/drivers/of/address.c
> index 7e59283a4472..df854bb427ce 100644
> --- a/drivers/of/address.c
> +++ b/drivers/of/address.c
> @@ -198,6 +198,25 @@ static u64 of_bus_pci_map(__be32 *addr, const __be32 *range, int na, int ns,
>  
>  #endif /* CONFIG_PCI */
>  
> +static int __of_address_resource_bounds(struct resource *r, u64 start, u64 size)
> +{
> +	u64 end = start;
> +
> +	if (overflows_type(start, r->start))
> +		return -EOVERFLOW;
> +	if (size == 0)
> +		return -EOVERFLOW;
> +	if (check_add_overflow(end, size - 1, &end))
> +		return -EOVERFLOW;
> +	if (overflows_type(end, r->end))
> +		return -EOVERFLOW;
 
This breaks PCI on powerpc qemu. Part of the PCI probe reads a resource
that's zero sized, which used to succeed but now fails due to the size
check above.

The diff below fixes it for me.

It leaves r.end == r.start, which is fine in my case, because the code
only uses r.start.

And it seems more sane than the old code which would return
end = start - 1, for zero sized resources.

cheers


diff --git a/drivers/of/address.c b/drivers/of/address.c
index df854bb427ce..a001e789a6c4 100644
--- a/drivers/of/address.c
+++ b/drivers/of/address.c
@@ -204,9 +204,7 @@ static int __of_address_resource_bounds(struct resource *r, u64 start, u64 size)
 
 	if (overflows_type(start, r->start))
 		return -EOVERFLOW;
-	if (size == 0)
-		return -EOVERFLOW;
-	if (check_add_overflow(end, size - 1, &end))
+	if (size > 0 && check_add_overflow(end, size - 1, &end))
 		return -EOVERFLOW;
 	if (overflows_type(end, r->end))
 		return -EOVERFLOW;