Thread (89 messages) 89 messages, 6 authors, 2025-01-22

Re: [PATCH v9 10/27] resource: harden resource_contains

flat view

From: Alejandro Lucero Palau <hidden>
Date: 2025-01-20 16:10:28
Also in: linux-cxl

On 1/18/25 02:03, Dan Williams wrote:
alejandro.lucero-palau@ wrote:
quoted
From: Alejandro Lucero <redacted>

While resource_contains checks for IORESOURCE_UNSET flag for the
resources given, if r1 was initialized with 0 size, the function
returns a false positive. This is so because resource start and
end fields are unsigned with end initialised to size - 1 by current
resource macros.

Make the function to check for the resource size for both resources
since r2 with size 0 should not be considered as valid for the function
purpose.

Signed-off-by: Alejandro Lucero <redacted>
Suggested-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Jonathan Cameron <redacted>
---
  include/linux/ioport.h | 2 ++
  1 file changed, 2 insertions(+)
diff --git a/include/linux/ioport.h b/include/linux/ioport.h
index 5385349f0b8a..7ba31a222536 100644
--- a/include/linux/ioport.h
+++ b/include/linux/ioport.h
@@ -296,6 +296,8 @@ static inline unsigned long resource_ext_type(const struct resource *res)
  /* True iff r1 completely contains r2 */
  static inline bool resource_contains(const struct resource *r1, const struct resource *r2)
  {
+	if (!resource_size(r1) || !resource_size(r2))
+		return false;
I just worry that some code paths expect the opposite, that it is ok to
pass zero size resources and get a true result.

That is an interesting point, I would say close to philosophic arguments. I guess you mean the zero size resource being the one that is contained inside the non-zero one, because the other option is making my vision blurry. In fact, even that one makes me feel trapped in a window-less room, in summer, with a bunch of economists, I mean philosophers, and my phone without signal for emergency calls.


But maybe it is just  my lack of understanding and there exists a good reason for this possibility.


Bjorn, I guess the ball is in your side ...
Did you audit existing callers?
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help