Fix resource leak: Variable f going out of scope leaks the storage it points to

Tracked-On: OAM-108693
Signed-off-by: Zheng, XianjunX <[email protected]>

external/PRIVATE/libpciaccess: Klocwork Fixes

BZ: 119061

KW fixes for libpciaccess module

Change-Id: I334290fc1de5fbcfe56721feb27663e1a1518cd4
Signed-off-by: arsharma <[email protected]

external/PRIVATE/libpciaccess: Klocwork Fixes

BZ: 119061

KW fixes for libpciaccess module

Change-Id: I334290fc1de5fbcfe56721feb27663e1a1518cd4
Signed-off-by: arsharma <[email protected]>
Reviewed-on: http://android.intel.com:8080/115795
Reviewed-by: Satyanantha, Rama Gopal M <[email protected]>
Reviewed-by: Lister, Ian <[email protected]>
Reviewed-by: S, Deepak <[email protected]>
Reviewed-by: Barbalho, Rafael <[email protected]>
Reviewed-by: Reese, Armin C <[email protected]>
Reviewed-by: Stimson, Dale B <[email protected]>
Reviewed-by: Goel, Akash <[email protected]>
Tested-by: Stimson, Dale B <[email protected]>
Reviewed-by: buildbot <[email protected]>
Tested-by: buildbot <[email protected]>

show more ...

Don't include config.h on Android

Android doesn't use autotools, so the header is never generated.

A previous commit removed some of the includes; this commit hopefully
removes the rest.

Signed-of

Don't include config.h on Android

Android doesn't use autotools, so the header is never generated.

A previous commit removed some of the includes; this commit hopefully
removes the rest.

Signed-off-by: Chad Versace <[email protected]>

show more ...

Don't include config.h on Android

Android doesn't use autotools, so the header is never generated.

Signed-off-by: Chad Versace <[email protected]>

delete_io_handle: fix deletion of last handle

When num_ios goes from 1 to 0, a realloc(ios, 0); call is made.
This is equivalent to free(ios) and NULL is returned.

However, the previous logic in th

delete_io_handle: fix deletion of last handle

When num_ios goes from 1 to 0, a realloc(ios, 0); call is made.
This is equivalent to free(ios) and NULL is returned.

However, the previous logic in the code incorrectly discards this NULL
return value. When we next call new_io_handle(), realloc(ios, X) is
called with "ios" pointing to freed memory. This causes glibc to abort.

Correct this logic to detect the 1-to-0 case and handle it correctly.
Other cases are unchanged; there is still value in checking the
return value from realloc() as it also returns NULL on error.

Signed-off-by: Daniel Drake <[email protected]>
Reviewed-by: Alan Coopersmith <[email protected]>

show more ...

Add VGA Arbiter support for OpenBSD.

Signed-off-by: Matthieu Herrb <[email protected]>

OpenBSD: Indicate that devices need access to legacy VGA resources.

Makes sure xserver disables DRI if the VGA arbiter is in use.

Signed-off-by: Matthieu Herrb <[email protected]>

Solaris: improve support for sparc platform

(1) added prom property retrieval on sparc
(2) added multiple domain support on sparc
(3) use kernel device as mapping device
(4) performance improvements

Solaris: improve support for sparc platform

(1) added prom property retrieval on sparc
(2) added multiple domain support on sparc
(3) use kernel device as mapping device
(4) performance improvements by removing redundant ioctl

Signed-off-by: Henry Zhao <[email protected]>
Reviewed-by: Alan Coopersmith <[email protected]>
Signed-off-by: Alan Coopersmith <[email protected]>

show more ...

libpciaccess: close mtrr fd on pci_cleanup

Since the fd is not closed, calling pci_system_init and
pci_system_cleanup more than 1024 times results in "too many files open"
error.

Signed-off-by: Nit

libpciaccess: close mtrr fd on pci_cleanup

Since the fd is not closed, calling pci_system_init and
pci_system_cleanup more than 1024 times results in "too many files open"
error.

Signed-off-by: Nithin Nayak Sujir <[email protected]>
Reviewed-by: Jeremy Huddleston <[email protected]>

show more ...

Solaris: Give better error on realloc failure

commit a18460b385ae03 converted from a fixed maximum number of devices
to dynamically growing the list via realloc, but didn't update the
error message

Solaris: Give better error on realloc failure

commit a18460b385ae03 converted from a fixed maximum number of devices
to dynamically growing the list via realloc, but didn't update the
error message shown on failure.

Signed-off-by: Alan Coopersmith <[email protected]>
Reviewed-by: Dirk Wallenstein <[email protected]>
Reviewed-by: Mark Kettenis <[email protected]>

show more ...

linux sysfs: Fix read-write access in map_legacy

O_RDONLY | O_WRONLY != O_RDWR

><

Reported-by: Javier Pello <[email protected]>
Signed-off-by: Jeremy Huddleston <[email protected]>

Update library version to reflect new API

Signed-off-by: Jeremy Huddleston <[email protected]>

linux: Implement map_legacy

Signed-off-by: Adam Jackson <[email protected]>
Reviewed-by: Jeremy Huddleston <[email protected]>

Add map_legacy interface

This allows platforms to hand back mmaps of the low 1M (ISA) address
space on a per-domain basis.

Signed-off-by: Adam Jackson <[email protected]>
Reviewed-by: Jeremy Huddlest

Add map_legacy interface

This allows platforms to hand back mmaps of the low 1M (ISA) address
space on a per-domain basis.

Signed-off-by: Adam Jackson <[email protected]>
Reviewed-by: Jeremy Huddleston <[email protected]>

show more ...

Strip trailing whitespace

Performed with: find * -type f | xargs perl -i -p -e 's{[ \t]+$}{}'
git diff -w & git diff -b show no diffs from this change

Signed-off-by: Alan Coopersmith <alan.coopersm

Strip trailing whitespace

Performed with: find * -type f | xargs perl -i -p -e 's{[ \t]+$}{}'
git diff -w & git diff -b show no diffs from this change

Signed-off-by: Alan Coopersmith <[email protected]>

show more ...

linux: Fix a crash in populate_devices

If scandir returns -1, the 'devices' array won't be initialized, and
attempting to free() it will crash.

Signed-off-by: Adam Jackson <[email protected]>

linux: Only set errno after an error

errno is only valid after an error, and was being filled with a garbage
value upon eof.

Signed-off-by: Chris Wilson <[email protected]>

Solaris support for multiple PCI segments (domains)

1. Removed hardcoded maximum size of 256 PCI devices, which is
too small for large systems. The number of devices is dynamically
resized as neede

Solaris support for multiple PCI segments (domains)

1. Removed hardcoded maximum size of 256 PCI devices, which is
too small for large systems. The number of devices is dynamically
resized as needed.

2. pci_device_solx_devfs_probe() no longer walks the device tree
from the very top ("/") but instead starts at the nexus which
owns the bus. Performance optimization for systems with multiple
bus nodes (including systems with just one segment/domain).

3. Added support for multiple domains/segments. Code tested
on kernels with and without multiple segment support so it should
be safe to integrate independent of the kernel version.

Signed-off-by: John Martin <[email protected]>
Signed-off-by: Alan Coopersmith <[email protected]>
Reviewed-by: Adam Jackson <[email protected]>

show more ...

Fix detection of multi-function cards

Bit 7 of the pci device header type is an indicator of multi-function
devices, so needs to be masked when examining the bridge type.
Thanks go to Chih-Pin Wu fo

Fix detection of multi-function cards

Bit 7 of the pci device header type is an indicator of multi-function
devices, so needs to be masked when examining the bridge type.
Thanks go to Chih-Pin Wu for finding the bug and its fix.

https://bugs.launchpad.net/ubuntu/+source/libpciaccess/+bug/681207

Reviewed-by: Adam Jackson <[email protected]>
Signed-off-by: Bryce Harrington <[email protected]>

show more ...

Revert "Don't try to use bogus bridge data"

This reverts commit 2bda5b733bb12854760750c08138db95e77aea0c.

It actually conflicts with 9ba94caf57e3a8c3e9c6f3f5f068f4a7a7b3ff9d
which came earlier, but

Revert "Don't try to use bogus bridge data"

This reverts commit 2bda5b733bb12854760750c08138db95e77aea0c.

It actually conflicts with 9ba94caf57e3a8c3e9c6f3f5f068f4a7a7b3ff9d
which came earlier, but I didn't catch it probably because I did a blind
rebase before pushing.

When reading bridge bus info, we may need to read out the current bridge
info if the PCI bridge private is uninitialized, so remove the check at
the top that would always short circuit the case where the priv was
still 0.

References:
https://bugs.launchpad.net/ubuntu/+source/libpciaccess/+bug/681207.

Reported-by: Bryce Harrington <[email protected]>
Signed-off-by: Jesse Barnes <[email protected]>

show more ...

Sun's copyrights belong to Oracle now

Signed-off-by: Alan Coopersmith <[email protected]>

include: add a Makefile to install and distribute header file

Makefiles are simpler when they only handle what is in their
directory.

Reviewed-by: Rémi Cardona <[email protected]>
Signed-off-by: Gaet

include: add a Makefile to install and distribute header file

Makefiles are simpler when they only handle what is in their
directory.

Reviewed-by: Rémi Cardona <[email protected]>
Signed-off-by: Gaetan Nadon <[email protected]>

show more ...

config: replace INCLUDES, which is deprecated, with AM_CPPFLAGS

Fixes an Automake warning.

Reviewed-by: Rémi Cardona <[email protected]>
Signed-off-by: Gaetan Nadon <[email protected]>

config: provide a makefile for scanpci man pages

Reviewed-by: Rémi Cardona <[email protected]>
Signed-off-by: Gaetan Nadon <[email protected]>

config: provide a makefile for scanpci dev tool

No function changes.

Note that automake generates a wrapper script if libpciaccess
is not installed at configuration time, which is the most common c

config: provide a makefile for scanpci dev tool

No function changes.

Note that automake generates a wrapper script if libpciaccess
is not installed at configuration time, which is the most common case.

If it is installed at configuration time, you get an executable.

The makefile build order ensures the lib is build before scanpci.

Reviewed-by: Rémi Cardona <[email protected]>
Signed-off-by: Gaetan Nadon <[email protected]>

show more ...