video/aperture: match the pci device when calling sysfb_disable()
PATCH v2 1/1 video/aperture: match the pci device when calling sysfb_disable()
In aperture_remove_conflicting_pci_devices(), we currently only call sysfb_disable() on vga class devices. This leads to the following problem when the pimary device is not VGA compatible:
1. A PCI device with a non-VGA class is the boot display
2. That device is probed first and it is not a VGA device so sysfb_disable() is not called, but the device resources are freed by aperture_detach_platform_device()
3. Non-primary GPU has a VGA class and it ends up calling sysfb_disable()
4. NULL pointer dereference via sysfb_disable() since the resources have already been freed by aperture_detach_platform_device() when it was called by the other device.
Fix this by passing a device pointer to sysfb_disable() and checking the device to determine if we should execute it or not.
v2: Fix build when CONFIG_SCREEN_INFO is not set
--- drivers/firmware/sysfb.c | 11 +++++++++-- drivers/of/platform.c | 2 +- drivers/video/aperture.c | 5 ++--- include/linux/sysfb.h | 4 ++-- 4 files changed, 14 insertions(+), 8 deletions(-)
drivers/firmware/sysfb.c +9 -2
@@ -39,6 +39,8 @@ static struct platform_device *pd;static DEFINE_MUTEX (disable_lock);static bool disabled; +static struct device *sysfb_parent_dev (const struct screen_info *si); +static bool sysfb_unregister (void ) {if (IS_ERR_OR_NULL (pd))
@@ -52,6 +54,7 @@ static bool sysfb_unregister(void)/** * sysfb_disable() - disable the Generic System Framebuffers support +* @dev: the device to check if non-NULL * * This disables the registration of system framebuffer devices that match the * generic drivers that make use of the system framebuffer set up by firmware.
@@ -61,8 +64,12 @@ static bool sysfb_unregister(void)* Context: The function can sleep. A @disable_lock mutex is acquired to serialize * against sysfb_init(), that registers a system framebuffer device. */ -void sysfb_disable (void ) +void sysfb_disable (struct device *dev) { +struct screen_info *si = &screen_info; + +if (dev && dev !=sysfb_parent_dev (si)) +return ;mutex_lock (&disable_lock);sysfb_unregister (); disabled = true;
@@ -93,7 +100,7 @@ static __init bool sysfb_pci_dev_is_enabled(struct pci_dev *pdev) }#endif -static __init struct device *sysfb_parent_dev (const struct screen_info *si) +static struct device *sysfb_parent_dev (const struct screen_info *si) {struct pci_dev *pdev;
drivers/of/platform.c +1 -1
@@ -592,7 +592,7 @@ static int __init of_platform_default_populate_init(void)* This can happen for example on DT systems that do EFI * booting and may provide a GOP handle to the EFI stub. */ -sysfb_disable (); +sysfb_disable (NULL );of_platform_device_create (node,NULL ,NULL );of_node_put (node); }
drivers/video/aperture.c +2 -3
@@ -293,7 +293,7 @@ int aperture_remove_conflicting_devices(resource_size_t base, resource_size_t si* ask for this, so let's assume that a real driver for the display * was already probed and prevent sysfb to register devices later. */ -sysfb_disable (); +sysfb_disable (NULL );aperture_detach_devices (base, size);
@@ -353,8 +353,7 @@ int aperture_remove_conflicting_pci_devices(struct pci_dev *pdev, const char *naif (pdev ==vga_default_device ()) primary = true; -if (primary) -sysfb_disable (); +sysfb_disable (&pdev->dev);for (bar =0 ; bar <PCI_STD_NUM_BARS ; ++bar) {if (!(pci_resource_flags (pdev, bar) &IORESOURCE_MEM ))
include/linux/sysfb.h +2 -2
@@ -58,11 +58,11 @@ struct efifb_dmi_info {#ifdef CONFIG_SYSFB -void sysfb_disable (void ); +void sysfb_disable (struct device *dev);#else /* CONFIG_SYSFB */ -static inline void sysfb_disable (void ) +static inline void sysfb_disable (struct device *dev) { }
Signature · 1 line
2.46.0
I forgot to update the patch title but it should probably be something like:
video/aperture: optionally match the device in sysfb_disable()
Alex
Quoted 135 lines · Alex Deucher
In
aperture_remove_conflicting_pci_devices(), we currently only callsysfb_disable()on vga class devices. This leads to the following problem when the pimary device is not VGA compatible:1. A PCI device with a non-VGA class is the boot display
2. That device is probed first and it is not a VGA device sosysfb_disable()is not called, but the device resources are freed byaperture_detach_platform_device()
3. Non-primary GPU has a VGA class and it ends up callingsysfb_disable()
4. NULL pointer dereference viasysfb_disable()since the resources have already been freed byaperture_detach_platform_device()when it was called by the other device.Fix this by passing a device pointer to
sysfb_disable()and checking the device to determine if we should execute it or not.v2: Fix build when
CONFIG_SCREEN_INFOis not setFixes 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")Cc Javier Martinez Canillas <[email protected]>Cc Thomas Zimmermann <[email protected]>Cc Helge Deller <[email protected]>Cc Sam Ravnborg <[email protected]>Cc Daniel Vetter <[email protected]>Signed-off-by Alex Deucher <[email protected]>--- drivers/firmware/sysfb.c | 11 +++++++++-- drivers/of/platform.c | 2 +- drivers/video/aperture.c | 5 ++--- include/linux/sysfb.h | 4 ++-- 4 files changed, 14 insertions(+), 8 deletions(-)drivers/firmware/sysfb.c +9
-2@@ -39,6 +39,8 @@ static struct platform_device *pd;static DEFINE_MUTEX (disable_lock);static bool disabled; +static struct device *sysfb_parent_dev (const struct screen_info *si); +static bool sysfb_unregister (void ) {if (IS_ERR_OR_NULL (pd))@@ -52,6 +54,7 @@ static bool sysfb_unregister(void)/** * sysfb_disable() - disable the Generic System Framebuffers support +* @dev: the device to check if non-NULL * * This disables the registration of system framebuffer devices that match the * generic drivers that make use of the system framebuffer set up by firmware. @@ -61,8 +64,12 @@ static bool sysfb_unregister(void)* Context: The function can sleep. A @disable_lock mutex is acquired to serialize * against sysfb_init(), that registers a system framebuffer device. */ -void sysfb_disable (void ) +void sysfb_disable (struct device *dev) { +struct screen_info *si = &screen_info; + +if (dev && dev !=sysfb_parent_dev (si)) +return ;mutex_lock (&disable_lock);sysfb_unregister (); disabled = true;@@ -93,7 +100,7 @@ static __init bool sysfb_pci_dev_is_enabled(struct pci_dev *pdev) }#endif -static __init struct device *sysfb_parent_dev (const struct screen_info *si) +static struct device *sysfb_parent_dev (const struct screen_info *si) {struct pci_dev *pdev;drivers/of/platform.c +1
-1@@ -592,7 +592,7 @@ static int __init of_platform_default_populate_init(void)* This can happen for example on DT systems that do EFI * booting and may provide a GOP handle to the EFI stub. */ -sysfb_disable (); +sysfb_disable (NULL );of_platform_device_create (node,NULL ,NULL );of_node_put (node); }drivers/video/aperture.c +2
-3@@ -293,7 +293,7 @@ int aperture_remove_conflicting_devices(resource_size_t base, resource_size_t si* ask for this, so let's assume that a real driver for the display * was already probed and prevent sysfb to register devices later. */ -sysfb_disable (); +sysfb_disable (NULL );aperture_detach_devices (base, size);@@ -353,8 +353,7 @@ int aperture_remove_conflicting_pci_devices(struct pci_dev *pdev, const char *naif (pdev ==vga_default_device ()) primary = true; -if (primary) -sysfb_disable (); +sysfb_disable (&pdev->dev);for (bar =0 ; bar <PCI_STD_NUM_BARS ; ++bar) {if (!(pci_resource_flags (pdev, bar) &IORESOURCE_MEM ))include/linux/sysfb.h +2
-3@@ -58,11 +58,11 @@ struct efifb_dmi_info {#ifdef CONFIG_SYSFB -void sysfb_disable (void ); +void sysfb_disable (struct device *dev);#else /* CONFIG_SYSFB */ -static inline void sysfb_disable (void ) +static inline void sysfb_disable (struct device *dev) { } --2.46.0
Alex Deucher <[email protected]> writes:
Hello Alex,
Quoted 26 lines
In
aperture_remove_conflicting_pci_devices(), we currently only callsysfb_disable()on vga class devices. This leads to the following problem when the pimary device is not VGA compatible:1. A PCI device with a non-VGA class is the boot display
2. That device is probed first and it is not a VGA device sosysfb_disable()is not called, but the device resources are freed byaperture_detach_platform_device()
3. Non-primary GPU has a VGA class and it ends up callingsysfb_disable()
4. NULL pointer dereference viasysfb_disable()since the resources have already been freed byaperture_detach_platform_device()when it was called by the other device.Fix this by passing a device pointer to
sysfb_disable()and checking the device to determine if we should execute it or not.v2: Fix build when
CONFIG_SCREEN_INFOis not setFixes 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")Cc Javier Martinez Canillas <[email protected]>Cc Thomas Zimmermann <[email protected]>Cc Helge Deller <[email protected]>Cc Sam Ravnborg <[email protected]>Cc Daniel Vetter <[email protected]>Signed-off-by Alex Deucher <[email protected]>
Quoted 1 line
---
The patch looks good to me.
I just have to minor comments below:
...
Quoted 7 lines
/**
*sysfb_disable()- disable the Generic System Framebuffers support
+ * @dev: the device to check if non-NULL
*
* This disables the registration of system framebuffer devices that match the
* generic drivers that make use of the system framebuffer set up by firmware.
drivers/firmware/sysfb.c static bool sysfb_unregister(void) · line 72
Quoted 10 lines
drivers/firmware/sysfb.c +5
-1@@ -61,8 +64,12 @@ static bool sysfb_unregister(void)* Context: The function can sleep. A @disable_lock mutex is acquired to serialize * against sysfb_init(), that registers a system framebuffer device. */ -void sysfb_disable (void ) +void sysfb_disable (struct device *dev) { +struct screen_info *si = &screen_info; + +if (dev && dev !=sysfb_parent_dev (si)) +return ;
Does this need to be protected by the disable_lock mutex? i.e:
mutex_lock(&disable_lock);
if (!dev || dev == sysfb_parent_dev(si) {
sysfb_unregister();
disabled = true;
}
mutex_unlock(&disable_lock);...
drivers/video/aperture.c int aperture_remove_conflicting_pci_devices(struct pci_dev *pdev, const char *na · line 357
Quoted 8 lines
drivers/video/aperture.c +1
-2@@ -353,8 +353,7 @@ int aperture_remove_conflicting_pci_devices(struct pci_dev *pdev, const char *naif (pdev ==vga_default_device ()) primary = true; -if (primary) -sysfb_disable (); +sysfb_disable (&pdev->dev);
After this change the primary variable is only used to determine whether __aperture_remove_legacy_vga_devices(pdev) should be called or not. So I wonder if could just be dropped and instead have:
/* * If this is the primary adapter, there could be a VGA device * that consumes the VGA framebuffer I/O range. Remove this * device as well. */ if (pdev == vga_default_device()) ret = __aperture_remove_legacy_vga_devices(pdev);
Signature · 5 lines
Best regards, Javier Martinez Canillas Core Platforms Red Hat
Patchwork
CI✗ Fi.CI.BAT: failure for video/aperture: match the pci device when calling sysfb_disable() (rev2)
In reply to Alex Deucher== Series Details ==
== Series Details ==
Series: video/aperture: match the pci device when calling sysfb_disable() (rev2)
URL : https://patchwork.freedesktop.org/series/137088/
State : failure
== Summary ==
CI Bug Log - changes from CI_DRM_15261 -> Patchwork_137088v2
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with Patchwork_137088v2 absolutely need to be verified manually.
If you think the reported changes have nothing to do with the changes
introduced in Patchwork_137088v2, please notify your bug team ([email protected]) to allow them to document this new failure mode, which will reduce false positives in CI.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_137088v2/index.html
Participating hosts (39 -> 36)
------------------------------
Missing (3): fi-kbl-8809g fi-snb-2520m fi-elk-e7500
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in Patchwork_137088v2:
### IGT changes ###
#### Possible regressions ####
* igt@fbdev@eof:
- bat-arls-1: [PASS][1] -> [DMESG-WARN][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_15261/bat-arls-1/igt@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_137088v2/bat-arls-1/igt@[email protected]
Build changes
-------------
* Linux: CI_DRM_15261 -> Patchwork_137088v2
CI-20190529: 20190529
CI_DRM_15261: 19d5b3fcc4a8dcf26076e522c794164f4d3e40c0 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_7975: 7975
Patchwork_137088v2: 19d5b3fcc4a8dcf26076e522c794164f4d3e40c0 @ git://anongit.freedesktop.org/gfx-ci/linux
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_137088v2/index.html
kernel test robot
build botkernel test robot noticed the following build warnings:
Hi Alex,
kernel test robot noticed the following build warnings:
[auto build test WARNING on drm-tip/drm-tip]
[also build test WARNING on robh/for-next drm-misc/drm-misc-next linus/master v6.11-rc4 next-20240820] [If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Alex-Deucher/video-aperture-match-the-pci-device-when-calling-sysfb_disable/20240820-005528 base: git://anongit.freedesktop.org/drm/drm-tip drm-tip patch link: https://lore.kernel.org/r/20240819165341.799848-1-alexander.deucher%40amd.com patch subject: [PATCH V2] video/aperture: match the pci device when calling sysfb_disable() config: i386-randconfig-054-20240820 (https://download.01.org/0day-ci/archive/20240821/[email protected]/config) compiler: clang version 18.1.5 (https://github.com/llvm/llvm-project 617a15a9eac96088ae5e9134248d8236e34b91b1) reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240821/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All warnings (new ones prefixed by >>, old ones prefixed by <<):
>> WARNING: modpost: vmlinux: section mismatch in reference: sysfb_disable+0x7a (section: .text) -> sysfb_pci_dev_is_enabled (section: .init.text)
WARNING: modpost: missing MODULE_DESCRIPTION() in lib/test_objpool.o
WARNING: modpost: missing MODULE_DESCRIPTION() in lib/asn1_decoder.o
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki