From: Marcelo Diop-Gonzalez <marcgonzalez@google.com>
To: gregkh@linuxfoundation.org, eric@anholt.net, wahrenst@gmx.net
Cc: devel@driverdev.osuosl.org, linux-rpi-kernel@lists.infradead.org
Subject: [PATCH 0/3] staging: vchiq: Fix vchiq_read return value in case of error
Date: Wed, 20 Nov 2019 10:36:45 -0500 [thread overview]
Message-ID: <cover.1574253964.git.marcgonzalez@google.com> (raw)
This is a proposed fix of an issue regarding the handling of positive
return values from copy_to_user() in vchiq_read(). It looks like the
struct dump_context passed to the vchiq_dump() function keeps track
of the number of bytes written in the context->actual field. When
copy_to_user() returns a positive value, this is set to -EFAULT. The
problem is that this is never returned to the user, and instead the function
continues, adding the number of bytes that should have been copied
to context->actual.
Running the following program:
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <unistd.h>
#include <errno.h>
int main(int argc, char **argv) {
int fd = open("/dev/vchiq", 0);
if (fd < 0)
exit(-1);
sbrk(getpagesize());
char *bad_buf = sbrk(0)-100;
int x = read(fd, bad_buf, 2000);
printf("%d %d\n", x, errno);
puts(bad_buf);
}
gives this output:
-1 1
State 0: CONNECTED
tx_po
Remote: slots 2-32 tx_pos=578 recycle=1f
Slots claimed:
Note the mangled output and incorrect errno value. Messing with the
constants in that toy program changes the results. Sometimes read()
returns with no error, sometimes it returns with a wrong error code,
sometimes with the right one. But it seems that it only ever returns an
error at all accidentally, due to the fact that the comparison between
context->actual and context->space in vchiq_dump() is unsigned, so that
that function won't do anything else if it ever sets context->actual
to a negative value.
After this patchset, the above program prints this:
-1 14
State 0: CONNECTED
tx_pos=b4a218(@165de6b4), rx_pos=ae0668(@f02b54f4)
Version: 8 (min 3)
Stats
Help with testing would be appreciated. So far I've basically just
diffed the output of 'cat /dev/vchiq', run the program above with
a few different values, and run vchiq_test a few times.
These were applied to the staging-next branch of the tree
at git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging.git
Marcelo Diop-Gonzalez (3):
staging: vchiq_dump: Replace min with min_t
staging: vchiq: Refactor indentation in vchiq_dump_* functions
staging: vchiq: Have vchiq_dump_* functions return an error code
.../interface/vchiq_arm/vchiq_2835_arm.c | 5 +-
.../interface/vchiq_arm/vchiq_arm.c | 129 ++++++++++--------
.../interface/vchiq_arm/vchiq_core.c | 104 ++++++++++----
.../interface/vchiq_arm/vchiq_core.h | 12 +-
4 files changed, 153 insertions(+), 97 deletions(-)
--
2.20.1
_______________________________________________
devel mailing list
devel@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
next reply other threads:[~2019-11-20 16:27 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-11-20 15:36 Marcelo Diop-Gonzalez [this message]
2019-11-20 15:36 ` [PATCH 1/3] staging: vchiq_dump: Replace min with min_t Marcelo Diop-Gonzalez
2019-11-20 15:46 ` Greg KH
2019-11-20 15:36 ` [PATCH 2/3] staging: vchiq: Refactor indentation in vchiq_dump_* functions Marcelo Diop-Gonzalez
2019-11-20 15:36 ` [PATCH 3/3] staging: vchiq: Have vchiq_dump_* functions return an error code Marcelo Diop-Gonzalez
2019-11-20 18:50 ` Dan Carpenter
2019-11-21 21:25 ` Marcelo Diop-Gonzalez
2019-11-20 20:20 ` [PATCH v2 0/4] staging: vchiq: Fix vchiq_read return value in case of error Marcelo Diop-Gonzalez
2019-11-20 20:20 ` [PATCH v2 1/4] staging: vchiq: Fix block comment format in vchiq_dump() Marcelo Diop-Gonzalez
2019-11-20 20:21 ` [PATCH v2 2/4] staging: vchiq_dump: Replace min with min_t Marcelo Diop-Gonzalez
2019-11-20 20:21 ` [PATCH v2 3/4] staging: vchiq: Refactor indentation in vchiq_dump_* functions Marcelo Diop-Gonzalez
2019-11-20 20:21 ` [PATCH v2 4/4] staging: vchiq: Have vchiq_dump_* functions return an error code Marcelo Diop-Gonzalez
2019-11-21 19:38 ` [PATCH v2 0/4] staging: vchiq: Fix vchiq_read return value in case of error Nicolas Saenz Julienne
2019-11-21 21:24 ` Marcelo Diop-Gonzalez
2019-11-21 10:42 ` [PATCH 0/3] " Stefan Wahren
2019-11-21 19:41 ` Nicolas Saenz Julienne
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=cover.1574253964.git.marcgonzalez@google.com \
--to=marcgonzalez@google.com \
--cc=devel@driverdev.osuosl.org \
--cc=eric@anholt.net \
--cc=gregkh@linuxfoundation.org \
--cc=linux-rpi-kernel@lists.infradead.org \
--cc=wahrenst@gmx.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).