linux-kernel - [PATCH 5.19 0457/1157] media: atomisp: revert "dont pass a pointer to a local variable"

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: <20220815180457.875359702@linuxfoundation.org>
Date:   Mon, 15 Aug 2022 19:56:53 +0200
From:   Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To:     linux-kernel@...r.kernel.org
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        stable@...r.kernel.org, Hans de Goede <hdegoede@...hat.com>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Sasha Levin <sashal@...nel.org>
Subject: [PATCH 5.19 0457/1157] media: atomisp: revert "dont pass a pointer to a local variable"

From: Hans de Goede <hdegoede@...hat.com>

[ Upstream commit a3b36a8ce3d0c277fe243fa1be6bd3f606ed130f ]

The gcc is warning about returning a pointer to a local variable
is a false positive.

The type of handle is "struct ia_css_rmgr_vbuf_handle **" and
"h.vptr" is left to NULL, so the "if ((*handle)->vptr == 0x0)"
check always succeeds when the "*handle = &h;" statement which
gcc warns about executes. Leading to this statement being executed:

	rmgr_pop_handle(pool, handle);

If that succeeds,  then *handle has been set to point to one of
the pre-allocated array of handles, so it no longer points to h.

If that fails the following statement will be executed:

	/* Note that handle will change to an internally maintained one */
	ia_css_rmgr_refcount_retain_vbuf(handle);

Which allocated a new handle from the array of pre-allocated handles
and then makes *handle point to this. So the address of h is actually
never returned.

The fix for the false-postive compiler warning actually breaks the code,
the new:

	**handle = h;

is part of a "if (pool->copy_on_write) { ... }" which means that the
handle where *handle points to should be treated read-only, IOW
**handle must never be set, instead *handle must be set to point to
a new handle (with a copy of the contents of the old handle).

The old code correctly did this and the new fixed code gets this wrong.

Note there is another patch in this series, which fixes the warning
in another way.

Link: https://lore.kernel.org/linux-media/20220612160556.108264-2-hdegoede@redhat.com
Fixes: fa1451374ebf ("media: atomisp: don't pass a pointer to a local variable")
Signed-off-by: Hans de Goede <hdegoede@...hat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@...nel.org>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
 .../staging/media/atomisp/pci/runtime/rmgr/src/rmgr_vbuf.c    | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/media/atomisp/pci/runtime/rmgr/src/rmgr_vbuf.c b/drivers/staging/media/atomisp/pci/runtime/rmgr/src/rmgr_vbuf.c
index 39604752785b..d96aaa4bc75d 100644
--- a/drivers/staging/media/atomisp/pci/runtime/rmgr/src/rmgr_vbuf.c
+++ b/drivers/staging/media/atomisp/pci/runtime/rmgr/src/rmgr_vbuf.c
@@ -254,7 +254,7 @@ void rmgr_pop_handle(struct ia_css_rmgr_vbuf_pool *pool,
 void ia_css_rmgr_acq_vbuf(struct ia_css_rmgr_vbuf_pool *pool,
 			  struct ia_css_rmgr_vbuf_handle **handle)
 {
-	struct ia_css_rmgr_vbuf_handle h = { 0 };
+	struct ia_css_rmgr_vbuf_handle h;

 	if ((!pool) || (!handle) || (!*handle)) {
 		IA_CSS_LOG("Invalid inputs");
@@ -272,7 +272,7 @@ void ia_css_rmgr_acq_vbuf(struct ia_css_rmgr_vbuf_pool *pool,
 			h.size = (*handle)->size;
 			/* release ref to current buffer */
 			ia_css_rmgr_refcount_release_vbuf(handle);
-			**handle = h;
+			*handle = &h;
 		}
 		/* get new buffer for needed size */
 		if ((*handle)->vptr == 0x0) {
-- 
2.35.1