| Age | Commit message (Collapse) | Author |
|---|
|
[why]
this dscclk use DCN defined per DPM level will cause a DCFCLK increase.
needs to follow up.
This reverts commit 15b959534a39530a21d378190557cc8d1eab7b09
Reviewed-by: Yihan Zhu <yihan.zhu@amd.com>
Reviewed-by: Alvin Lee <alvin.lee2@amd.com>
Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Why]
Depending on when the HW latching event (vupdate) of double-buffered
registers happen relative to the PSR SDP (signals panel psr enter/exit)
deadline, and how bad the Panel clock has drifted since the last ALPM
off event, there can be up to 3 frames of delay between sending the PSR
exit cmd to DMUB fw, and when the panel starts displaying live frames.
This can manifest as micro-stuttering when userspace commit patterns
cause rapid toggling of the DRM vblank counter, since PSR enter/exit is
hooked up to DRM vblank disable/enable respectively.
In the ideal world, the panel should present the live frame immediately
on PSR exit cmd. But due to HW design and PSR limitations, immediate
exit can only happen by chance, when:
1. PSR exit cmd is ack'd by FW before HW latching (vupdate) event, and
2. Panel's SDP deadline -- determined by it's PSR Start Delay in DPCD
71h -- is after the vupdate event. The PSR exit SDP can then be sent
immediately after HW latches. Otherwise, we have to wait 1 frame. And
3. There is negligible drift between the panel's clock and source clock.
Otherwise, there can be up to 1 frame of drift.
Note that this delay is not expected with Panel Replay.
[How]
Since PSR power savings can be quite substantial, and there are a lot of
systems in the wild with PSR panels, It'll be nice to have a middle
ground that balances user experience with power savings.
A simple way to achieve this is by extending the vblank offdelay, such
that additional PSR exit delays will be less perceivable.
We can set:
20/100 * offdelay_ms = 3_frames_ms
=> offdelay_ms = 5 * 3_frames_ms
This ensures that `3_frames_ms` will only be experienced as a 20% delay
on top how long the panel has been static, and thus make the delay
less perceivable.
If this ends up being too high of a percentage, it can be dropped
further in a future change.
Fixes: 537ef0f88897 ("drm/amd/display: use new vblank enable policy for DCN35+")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Leo Li <sunpeng.li@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
If compiled without SI or CIK support but amdgpu tries to load it
will run into failures with uninitialized callbacks.
Show a nicer message in this case and fail probe instead.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4050
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
Hook up zero RPM enable for 9070 and 9070 XT based on RDNA3
(smu 13.0.0 and 13.0.7) code.
Tested on 9070 XT Hellhound
Signed-off-by: Tomasz Pakuła <tomasz.pakula.oficjalny@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.12.x
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
The user can set any speed value.
If speed is greater than UINT_MAX/8, division by zero is possible.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: c52dcf49195d ("drm/amd/pp: Avoid divide-by-zero in fan_ctrl_set_fan_speed_rpm")
Signed-off-by: Denis Arefev <arefev@swemel.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
The user can set any speed value.
If speed is greater than UINT_MAX/8, division by zero is possible.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: c52dcf49195d ("drm/amd/pp: Avoid divide-by-zero in fan_ctrl_set_fan_speed_rpm")
Signed-off-by: Denis Arefev <arefev@swemel.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
The user can set any speed value.
If speed is greater than UINT_MAX/8, division by zero is possible.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: 031db09017da ("drm/amd/powerplay/vega20: enable fan RPM and pwm settings V2")
Signed-off-by: Denis Arefev <arefev@swemel.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
The user can set any speed value.
If speed is greater than UINT_MAX/8, division by zero is possible.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: b64625a303de ("drm/amd/pm: correct the address of Arcturus fan related registers")
Signed-off-by: Denis Arefev <arefev@swemel.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
The user can set any speed value.
If speed is greater than UINT_MAX/8, division by zero is possible.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: c05d1c401572 ("drm/amd/swsmu: add aldebaran smu13 ip support (v3)")
Signed-off-by: Denis Arefev <arefev@swemel.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
|
|
Call device_remove_file() when driver remove.
Signed-off-by: Shixiong Ou <oushixiong@kylinos.cn>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Using device_create_with_groups() to simplify creation and removal.
Same as commit 1083a7be4504 ("tty: Use static attribute groups for
sysfs entries").
Signed-off-by: Shixiong Ou <oushixiong@kylinos.cn>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Added checks for xoffset, yoffset settings.
Incorrect settings of these parameters can lead to errors
in sm501fb_pan_ functions.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: 5fc404e47bdf ("[PATCH] fb: SM501 framebuffer driver")
Signed-off-by: Danila Chernetsov <listdansp@mail.ru>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Function dispc_ovl_setup is not intended to work with the value OMAP_DSS_WB
of the enum parameter plane.
The value of this parameter is initialized in dss_init_overlays and in the
current state of the code it cannot take this value so it's not a real
problem.
For the purposes of defensive coding it wouldn't be superfluous to check
the parameter value, because some functions down the call stack process
this value correctly and some not.
For example, in dispc_ovl_setup_global_alpha it may lead to buffer
overflow.
Add check for this value.
Found by Linux Verification Center (linuxtesting.org) with SVACE static
analysis tool.
Signed-off-by: Leonid Arapov <arapovl839@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Value of enum parameter 'plane' is initialized in dss_init_overlays and
cannot take the value OMAP_DSS_WB. Function dispc_ovl_setup_common could
be called with this value of parameter only from dispc_wb_setup, which has
never been used and has been removed in commit 4f55bb03801a
("omapfb: Remove unused writeback code"). The code in the if-branch is
unreachable.
Remove unreachable branch.
Found by Linux Verification Center (linuxtesting.org) with SVACE static
analysis tool.
Signed-off-by: Leonid Arapov <arapovl839@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Due to the nature of changes this is probably necessary. Even if these
drawing routines got way more testing than my patch submission scripts.
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
The original version duplicated more or less the same algorithms for
both system and i/o memory.
In this version the drawing algorithms (copy/fill/blit) are separate
from the memory access (system and i/o). The two parts are getting
combined in the loadable module sources. This also makes it more robust
against wrong memory access type or alignment mistakes as there's no
direct pointer access or arithmetic in the algorithm sources anymore.
Due to liberal use of inlining the compiled result is a single function
in all 6 cases, without unnecessary function calls. Unlike earlier the
use of macros could be minimized as apparently both gcc and clang is
capable now to do the same with inline functions just as well.
What wasn't quite the same in the two variants is the support for pixel
order reversing. This version is capable to do that for both system and
I/O memory, and not only for the latter. As demand for low bits per
pixel modes isn't high there's a configuration option to enable this
separately for the CFB and SYS modules.
The pixel reversing algorithm is different than earlier and was designed
so that it can take advantage of bit order reversing instructions on
architectures which have them. And even for higher bits per pixel modes
like four bpp.
One of the shortcomings of the earlier version was the incomplete
support for foreign endian framebuffers. Now all three drawing
algorithms produce correct output on both endians with native and
foreign framebuffers. This is one of the important differences even if
otherwise the algorithms don't look too different than before.
All three routines work now with aligned native word accesses. As a
consequence blitting isn't limited to 32 bits on 64 bit architectures as
it was before.
The old routines silently assumed that rows are a multiple of the word
size. Due to how the new routines function this isn't a requirement any
more and access will be done aligned regardless. However if the
framebuffer is configured like that then some of the fast paths won't be
available.
As this code is supposed to be running on all supported architectures it
wasn't optimized for a particular one. That doesn't mean I haven't
looked at the disassembly. That's where I noticed that it isn't a good
idea to use the fallback bitreversing code for example.
The low bits per pixel modes should be faster than before as the new
routines can blit 4 pixels at a time.
On the higher bits per pixel modes I retained the specialized aligned
routines so it should be more or less the same, except on 64 bit
architectures. There the blitting word size is double now which means 32
BPP isn't done a single pixel a time now.
The code was tested on x86, amd64, mips32 and mips64. The latter two in
big endian configuration. Originally thought I can get away with the
first two, but with such bit twisting code byte ordering is tricky and
not really possible to get right without actually verifying it.
While writing such routines isn't rocket science a lot of time was spent
on making sure that pixel ordering, foreign byte order, various bits per
pixels, cpu endianness and word size will give the expected result in
all sorts of combinations without making it overly complicated or full
with special cases.
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Remove dependency on fb_draw.h to allow fbcon packed pixel drawing refactoring.
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Remove dependency on fb_draw.h to allow fbcon packed pixel drawing refactoring.
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Use device_add_group() to simplify creation.
Signed-off-by: Shixiong Ou <oushixiong@kylinos.cn>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
[WHY]
1. The driver forgot to call device_remove_file()
in sh_mobile_lcdc_overlay_fb_unregister(), and there was
no error handling when calling device_create_file() failed.
2. This should probably use device_add_group() instead of
individual files to simplify both creation and removal. [Arnd]
3. The driver core can register and cleanup sysfs groups already.
as commit 95cdd538e0e5 ("fbdev: efifb: Register sysfs groups
through driver core").
[HOW]
Register sysfs groups through driver core.
Signed-off-by: Shixiong Ou <oushixiong@kylinos.cn>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
mdacon has roughly the same dependencies as vgacon but expresses them
as a negative list instead of a positive list, with the only practical
difference being PowerPC/CHRP, which uses vga16fb instead of vgacon.
The CONFIG_MDA_CONSOLE description advises to only turn it on when vgacon
is also used because MDA/Hercules-only systems should be using vgacon
instead, so just change the list to enforce that directly for simplicity.
The probing was broken from 2002 to 2008, this improves on the fix
that was added then: If vgacon is a loadable module, then mdacon
cannot be built-in now, and the list of systems that support vgacon
is carried over.
Fixes: 0b9cf3aa6b1e ("mdacon messing up default vc's - set default to vc13-16 again")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
dummycon fails to build on ARM/footbridge when the VGA console is
disabled, since I got the dependencies slightly wrong in a previous
patch:
drivers/video/console/dummycon.c: In function 'dummycon_init':
drivers/video/console/dummycon.c:27:25: error: 'CONFIG_DUMMY_CONSOLE_COLUMNS' undeclared (first use in this function); did you mean 'CONFIG_DUMMY_CONSOLE'?
27 | #define DUMMY_COLUMNS CONFIG_DUMMY_CONSOLE_COLUMNS
drivers/video/console/dummycon.c:28:25: error: 'CONFIG_DUMMY_CONSOLE_ROWS' undeclared (first use in this function); did you mean 'CONFIG_DUMMY_CONSOLE'?
28 | #define DUMMY_ROWS CONFIG_DUMMY_CONSOLE_ROWS
This only showed up after many thousand randconfig builds on Arm, and
doesn't matter in practice, but should still be fixed. Address it by
using the default row/columns on footbridge after all in that corner
case.
Fixes: 4293b0925149 ("dummycon: limit Arm console size hack to footbridge")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202409151512.LML1slol-lkp@intel.com/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Dummycon is used as a fallback conswitchp for vgacon and fbcon
in the VT code, and there are no references to it if all three
are disabled, so just leave it out of the kernel image for
configurations without those.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
The address of a data structure member was determined before
a corresponding null pointer check in the implementation of
the function “au1100fb_setmode”.
This issue was detected by using the Coccinelle software.
Fixes: 3b495f2bb749 ("Au1100 FB driver uplift for 2.6.")
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Acked-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
Convert to use devm_kmemdup() and devm_kmemdup_array() which are
more robust.
Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
The erase colour calculation for fbcon clearing should use get_color instead
of attr_col_ec, like everything else. The latter is similar but is not correct.
For example it's missing the depth dependent remapping and doesn't care about
blanking.
The problem can be reproduced by setting up the background colour to grey
(vt.color=0x70) and having an fbcon console set to 2bpp (4 shades of gray).
Now the background attribute should be 1 (dark gray) on the console.
If the screen is scrolled when pressing enter in a shell prompt at the bottom
line then the new line is cleared using colour 7 instead of 1. That's not
something fillrect likes (at 2bbp it expect 0-3) so the result is interesting.
This patch switches to get_color with vc_video_erase_char to determine the
erase colour from attr_col_ec. That makes the latter function redundant as
no other users were left.
Use correct erase colour for clearing in fbcon
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
I was wondering why there's garbage at the bottom of the screen when
tile blitting is used with an odd mode like 1080, 600 or 200. Sure there's
only space for half a tile but the same area is clean when the buffer
is bitmap.
Then later I found that it's supposed to be cleaned but that's not
implemented. So I took what's in bitblit and adapted it for tileblit.
This implementation was tested for both the horizontal and vertical case,
and now does the same as what's done for bitmap buffers.
If anyone is interested to reproduce the problem then I could bet that'd
be on a S3 or Ark. Just set up a mode with an odd line count and make
sure that the virtual size covers the complete tile at the bottom. E.g.
for 600 lines that's 608 virtual lines for a 16 tall tile. Then the
bottom area should be cleaned.
For the right side it's more difficult as there the drivers won't let an
odd size happen, unless the code is modified. But once it reports back a
few pixel columns short then fbcon won't use the last column. With the
patch that column is now clean.
Btw. the virtual size should be rounded up by the driver for both axes
(not only the horizontal) so that it's dividable by the tile size.
That's a driver bug but correcting it is not in scope for this patch.
Implement missing margin clearing for tileblit
Signed-off-by: Zsolt Kajtar <soci@c64.rulez.org>
Signed-off-by: Helge Deller <deller@gmx.de>
|
|
When arm64 is built with LTO, it upgrades READ_ONCE() to ldar / ldapr
(load-acquire) to avoid issues that can be caused by the compiler
optimizing away implicit address dependencies.
Unlike plain loads, these load-acquire instructions actually require an
aligned address.
For now, fix it by removing the READ_ONCE() that the buggy commit
introduced.
Fixes: ece69af2ede1 ("rwonce: handle KCSAN like KASAN in read_word_at_a_time()")
Reported-by: Nathan Chancellor <nathan@kernel.org>
Closes: https://lore.kernel.org/r/20250326203926.GA10484@ax162
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
|
|
Fix a silly bug where an array was used outside of its scope.
Fixes: 2051da858534 ("arm64/crc-t10dif: expose CRC-T10DIF function through lib")
Cc: stable@vger.kernel.org
Reported-by: David Binderman <dcb314@hotmail.com>
Closes: https://lore.kernel.org/r/AS8PR02MB102170568EAE7FFDF93C8D1ED9CA62@AS8PR02MB10217.eurprd02.prod.outlook.com
Link: https://lore.kernel.org/r/20250326200918.125743-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
|
|
Fix a silly bug where an array was used outside of its scope.
Fixes: 1684e8293605 ("arm/crc-t10dif: expose CRC-T10DIF function through lib")
Cc: stable@vger.kernel.org
Reported-by: David Binderman <dcb314@hotmail.com>
Closes: https://lore.kernel.org/r/AS8PR02MB102170568EAE7FFDF93C8D1ED9CA62@AS8PR02MB10217.eurprd02.prod.outlook.com
Link: https://lore.kernel.org/r/20250326200812.125574-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
|
|
A BPF scheduler may want to use the built-in idle cpumasks in ops.init()
before the scheduler is fully initialized, either directly or through a
BPF timer for example.
However, this would result in an error, since the idle state has not
been properly initialized yet.
This can be easily verified by modifying scx_simple to call
scx_bpf_get_idle_cpumask() in ops.init():
$ sudo scx_simple
DEBUG DUMP
===========================================================================
scx_simple[121] triggered exit kind 1024:
runtime error (built-in idle tracking is disabled)
...
Fix this by properly initializing the idle state before ops.init() is
called. With this change applied:
$ sudo scx_simple
local=2 global=0
local=19 global=11
local=23 global=11
...
Fixes: d73249f88743d ("sched_ext: idle: Make idle static keys private")
Signed-off-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Joel Fernandes <joelagnelf@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
create_dsq and therefore the scx_bpf_create_dsq kfunc currently silently
ignore duplicate entries. As a sched_ext scheduler is creating each DSQ
for a different purpose this is surprising behaviour.
Replace rhashtable_insert_fast which ignores duplicates with
rhashtable_lookup_insert_fast that reports duplicates (though doesn't
return their value). The rest of the code is structured correctly and
this now returns -EEXIST.
Tested by adding an extra scx_bpf_create_dsq to scx_simple. Previously
this was ignored, now init fails with a -17 code. Also ran scx_lavd
which continued to work well.
Signed-off-by: Jake Hillion <jake@hillion.co.uk>
Acked-by: Andrea Righi <arighi@nvidia.com>
Fixes: f0e1a0643a59 ("sched_ext: Implement BPF extensible scheduler class")
Cc: stable@vger.kernel.org # v6.12+
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull clocksource/event updates from Thomas Gleixner:
- Add support for suspend/resume in the STM32 LP-Timer driver with a
follow up fix, which uses the proper method to setup the timer as a
optional wakeup source instead of trying to force it as mandatory
wakeup source.
- The usual device tree updates to enable new SoC models in existing
drivers.
- Trivial spelling, style and indentation fixes
* tag 'timers-clocksource-2025-03-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
dt-bindings: timer: Add SiFive CLINT2
clocksource/drivers/stm32-lptimer: Use wakeup capable instead of init wakeup
clocksource/drivers/exynos_mct: Fixed a spelling error
clocksource/drivers/stm32-lptimer: Add support for suspend / resume
dt-bindings: timer: exynos4210-mct: add samsung,exynos2200-mct-peris compatible
dt-bindings: timer: exynos4210-mct: Add samsung,exynos990-mct compatible
dt-bindings: timer: Correct indentation and style in DTS example
|
|
Fixes "task out to lunch" warnings during recovery on large machines
with lots of dirty data in the journal.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
btree node scan has to wait on kthread workers that scan each device,
potentially for awhile.
We would like this to be interruptible, but we may need a different
mechanism than signals for that - we've had bugs in the past where
mounts were failing due to checking for signals, and no explanation on
where they came from.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Data move -> move_get_io_opts -> bch2_get_update_rebalance_opts
requires a not_extents iterator; this fixes the path where we're walking
the extents btree and chase a reflink pointer into the reflink btree.
bch2_lookup_indirect_extent() requires working with an extents iterator
(due to peek_slot() semantics), so we implement
bch2_lookup_indirect_extent_for_move().
This is simplified because there's no need to report
indirect_extent_missing_errors here, that can be deferred until fsck or
when a user reads that data.
Reported-by: Maël Kerbiriou <mael.kerbiriou@free.fr>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
There's various checks for "are we going to compress this" - but we're
not going to compress if we know it's incompressible.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
We weren't checking that accounting keys have the expected number of
accounters. Originally we probably wanted to be flexible on this, but it
doesn't look like that will be required - accounting is extended by
adding new counter types, not more counters to an existing type.
This means we can drop a BUG_ON() that popped once in automated testing,
and the new validation will make that bug easier to track down.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull MSI irq fix from Thomas Gleixner:
"An urgent fix for the XEN related PCI/MSI changes:
XEN used a global variable to disable the masking of MSI interrupts as
XEN handles that on the hypervisor side. This turned out to be a
problem with VMD as the PCI devices behind a VMD bridge are not always
handled by the hypervisor and then require masking by guest.
To solve this the global variable was replaced by a interrupt domain
specific flag, which is set by the generic XEN PCI/MSI domain, but not
by VMD or any other domain in the system.
So far, so good. But the implementation (and the reviewer) missed the
fact, that accessing the domain flag cannot be done directly because
there are at least two situations, where this fails.
Legacy architectures are not providing interrupt domains at all. The
new MSI parent domains do not require to have a domain info pointer.
Both cases result in a unconditional NULL pointer derefence.
The PCI/MSI code already has a function to query the MSI domain
specific flag in a safe way, which handles all possible cases of
PCI/MSI backends.
So the fix it simply to replace the open coded checks by invoking the
safe helper to query the flag"
* tag 'irq-urgent-2025-03-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
PCI/MSI: Handle the NOMASK flag correctly for all PCI/MSI backends
|
|
scx_select_cpu_dfl() has a meaningless conditional goto at the end. Remove
it. No functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andrea Righi <arighi@nvidia.com>
|
|
Return -EBUSY when using %SCX_PICK_IDLE_CORE with scx_select_cpu_dfl()
if a fully idle SMT core cannot be found, instead of falling back to
@prev_cpu, which is not a fully idle SMT core in this case.
Fixes: c414c2171cd9e ("sched_ext: idle: Honor idle flags in the built-in idle selection policy")
Signed-off-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
SMB1 protocol supports non-UNICODE (8-bit OEM character set) and
UNICODE (UTF-16) modes.
Linux SMB1 client implements both of them but currently does not allow to
choose non-UNICODE mode when SMB1 server announce UNICODE mode support.
This change adds a new mount option -o nounicode to disable UNICODE mode
and force usage of non-UNICODE (8-bit OEM character set) mode.
This allows to test non-UNICODE implementation of Linux SMB1 client against
any SMB1 server, including modern and recent Windows SMB1 server.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
Windows SMB servers (including SMB2+) which are working over RFC1001
require that Netbios server name specified in RFC1001 Session Request
packet is same as the UNC host name. Netbios server name can be already
specified manually via -o servern= option.
With this change the RFC1001 server name is set automatically by extracting
the hostname from the mount source.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
Commit ef7134c7fc48 ("smb: client: Fix use-after-free of network
namespace.") attempted to fix a netns use-after-free issue by manually
adjusting reference counts via sk->sk_net_refcnt and sock_inuse_add().
However, a later commit e9f2517a3e18 ("smb: client: fix TCP timers deadlock
after rmmod") pointed out that the approach of manually setting
sk->sk_net_refcnt in the first commit was technically incorrect, as
sk->sk_net_refcnt should only be set for user sockets. It led to issues
like TCP timers not being cleared properly on close. The second commit
moved to a model of just holding an extra netns reference for
server->ssocket using get_net(), and dropping it when the server is torn
down.
But there remain some gaps in the get_net()/put_net() balancing added by
these commits. The incomplete reference handling in these fixes results
in two issues:
1. Netns refcount leaks[1]
The problem process is as follows:
```
mount.cifs cifsd
cifs_do_mount
cifs_mount
cifs_mount_get_session
cifs_get_tcp_session
get_net() /* First get net. */
ip_connect
generic_ip_connect /* Try port 445 */
get_net()
->connect() /* Failed */
put_net()
generic_ip_connect /* Try port 139 */
get_net() /* Missing matching put_net() for this get_net().*/
cifs_get_smb_ses
cifs_negotiate_protocol
smb2_negotiate
SMB2_negotiate
cifs_send_recv
wait_for_response
cifs_demultiplex_thread
cifs_read_from_socket
cifs_readv_from_socket
cifs_reconnect
cifs_abort_connection
sock_release();
server->ssocket = NULL;
/* Missing put_net() here. */
generic_ip_connect
get_net()
->connect() /* Failed */
put_net()
sock_release();
server->ssocket = NULL;
free_rsp_buf
...
clean_demultiplex_info
/* It's only called once here. */
put_net()
```
When cifs_reconnect() is triggered, the server->ssocket is released
without a corresponding put_net() for the reference acquired in
generic_ip_connect() before. it ends up calling generic_ip_connect()
again to retry get_net(). After that, server->ssocket is set to NULL
in the error path of generic_ip_connect(), and the net count cannot be
released in the final clean_demultiplex_info() function.
2. Potential use-after-free
The current refcounting scheme can lead to a potential use-after-free issue
in the following scenario:
```
cifs_do_mount
cifs_mount
cifs_mount_get_session
cifs_get_tcp_session
get_net() /* First get net */
ip_connect
generic_ip_connect
get_net()
bind_socket
kernel_bind /* failed */
put_net()
/* after out_err_crypto_release label */
put_net()
/* after out_err label */
put_net()
```
In the exception handling process where binding the socket fails, the
get_net() and put_net() calls are unbalanced, which may cause the
server->net reference count to drop to zero and be prematurely released.
To address both issues, this patch ties the netns reference counting to
the server->ssocket and server lifecycles. The extra reference is now
acquired when the server or socket is created, and released when the
socket is destroyed or the server is torn down.
[1]: https://bugzilla.kernel.org/show_bug.cgi?id=219792
Fixes: ef7134c7fc48 ("smb: client: Fix use-after-free of network namespace.")
Fixes: e9f2517a3e18 ("smb: client: fix TCP timers deadlock after rmmod")
Signed-off-by: Wang Zhaolong <wangzhaolong1@huawei.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
cifs.ko is missing validation check when accessing smb_aces.
This patch add validation check for the fields in smb_aces.
Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
secondary channels.
In a multichannel setup, it was observed that a few fields were not being
copied over to the secondary channels, which impacted performance in cases
where these options were relevant but not properly synchronized. To address
this, this patch introduces copying the following parameters from the
primary channel to the secondary channels:
- min_offload
- compression.requested
- dfs_conn
- ignore_signature
- leaf_fullpath
- noblockcnt
- retrans
- sign
By copying these parameters, we ensure consistency across channels and
prevent performance degradation due to missing or outdated settings.
Cc: stable@vger.kernel.org
Signed-off-by: Aman <aman1@microsoft.com>
Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
Function ip_rfc1001_connect() send NetBIOS session request but currently
does not read response. It even does not wait for the response. Instead it
just calls usleep_range(1000, 2000) and explain in comment that some
servers require short break before sending SMB negotiate packet. Response
is later handled in generic is_smb_response() function called from
cifs_demultiplex_thread().
That comment probably refers to the old DOS SMB server which cannot process
incoming SMB negotiate packet if it has not sent NetBIOS session response
packet. Note that current sleep timeout is too small when trying to
establish connection to DOS SMB server running in qemu virtual machine
connected over qemu user networking with guestfwd netcat options. So that
usleep_range() call is not useful at all.
NetBIOS session response packet contains useful error information, like
the server name specified NetBIOS session request packet is incorrect.
Old Windows SMB servers and even the latest SMB server on the latest
Windows Server 2022 version requires that the name is the correct server
name, otherwise they return error RFC1002_NOT_PRESENT. This applies for all
SMB dialects (old SMB1, and also modern SMB2 and SMB3).
Therefore read the reply of NetBIOS session request and implement parsing
of the reply. Log received error to dmesg to help debugging reason why
connection was refused. Also convert NetBIOS error to useful errno.
Note that ip_rfc1001_connect() function is used only when doing connection
over port 139. So the common SMB scenario over port 445 is not affected by
this change at all.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
Function ip_rfc1001_connect() which establish NetBIOS session for SMB
connections, currently uses smb_send() function for sending NetBIOS Session
Request packet. This function expects that the passed buffer is SMB packet
and for SMB2+ connections it mangles packet header, which breaks prepared
NetBIOS Session Request packet. Result is that this function send garbage
packet for SMB2+ connection, which SMB2+ server cannot parse. That function
is not mangling packets for SMB1 connections, so it somehow works for SMB1.
Fix this problem and instead of smb_send(), use smb_send_kvec() function
which does not mangle prepared packet, this function send them as is. Just
API of this function takes struct msghdr (kvec) instead of packet buffer.
[MS-SMB2] specification allows SMB2 protocol to use NetBIOS as a transport
protocol. NetBIOS can be used over TCP via port 139. So this is a valid
configuration, just not so common. And even recent Windows versions (e.g.
Windows Server 2022) still supports this configuration: SMB over TCP port
139, including for modern SMB2 and SMB3 dialects.
This change fixes SMB2 and SMB3 connections over TCP port 139 which
requires establishing of NetBIOS session. Tested that this change fixes
establishing of SMB2 and SMB3 connections with Windows Server 2022.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
Currently ->get_acl() callback always create request for OWNER, GROUP and
DACL, even when only DACLs was requested by user. Change API callback to
request only information for which the caller asked. Therefore when only
DACLs requested, then SMB client will prepare and send DACL-only request.
This change fixes retrieving of "system.cifs_acl" and "system.smb3_acl"
xattrs to contain only DACL structure as documented.
Note that setting/changing of "system.cifs_acl" and "system.smb3_acl"
xattrs already takes only DACL structure and ignores all other fields.
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
