syzbot reminds us netlink_getname() runs locklessly [1]
This first patch annotates the race against nlk->portid.
Following patches take care of the remaining races.
[1]
BUG: KCSAN: data-race in netlink_getname / netlink_insert
write to 0xffff88814176d310 of 4 bytes by task 2315 on cpu 1:
netlink_insert+0xf1/0x9a0 net/netlink/af_netlink.c:583
netlink_autobind+0xae/0x180 net/netlink/af_netlink.c:856
netlink_sendmsg+0x444/0x760 net/netlink/af_netlink.c:1895
sock_sendmsg_nosec net/socket.c:714 [inline]
sock_sendmsg net/socket.c:734 [inline]
____sys_sendmsg+0x38f/0x500 net/socket.c:2476
___sys_sendmsg net/socket.c:2530 [inline]
__sys_sendmsg+0x19a/0x230 net/socket.c:2559
__do_sys_sendmsg net/socket.c:2568 [inline]
__se_sys_sendmsg net/socket.c:2566 [inline]
__x64_sys_sendmsg+0x42/0x50 net/socket.c:2566
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
read to 0xffff88814176d310 of 4 bytes by task 2316 on cpu 0:
netlink_getname+0xcd/0x1a0 net/netlink/af_netlink.c:1144
__sys_getsockname+0x11d/0x1b0 net/socket.c:2026
__do_sys_getsockname net/socket.c:2041 [inline]
__se_sys_getsockname net/socket.c:2038 [inline]
__x64_sys_getsockname+0x3e/0x50 net/socket.c:2038
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x2b/0x70 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
value changed: 0x00000000 -> 0xc9a49780
Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 2316 Comm: syz-executor.2 Not tainted 6.2.0-rc3-syzkaller-00030-ge8f60cd7db24-dirty #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/26/2022
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Skip interference with an ongoing transaction, do not perform garbage
collection on inactive elements. Reset annotated previous end interval
if the expired element is marked as busy (control plane removed the
element right before expiration).
...instead of a tree descent, which became overly complicated in an
attempt to cover cases where expired or inactive elements would affect
comparisons with the new element being inserted.
Further, it turned out that it's probably impossible to cover all those
cases, as inactive nodes might entirely hide subtrees consisting of a
complete interval plus a node that makes the current insertion not
overlap.
To speed up the overlap check, descent the tree to find a greater
element that is closer to the key value to insert. Then walk down the
node list for overlap detection. Starting the overlap check from
rb_first() unconditionally is slow, it takes 10 times longer due to the
full linear traversal of the list.
Moreover, perform garbage collection of expired elements when walking
down the node list to avoid bogus overlap reports.
For the insertion operation itself, this essentially reverts back to the
implementation before commit 4392e5177e05 ("netfilter: nft_set_rbtree:
Detect partial overlaps on insertion"), except that cases of complete
overlap are already handled in the overlap detection phase itself, which
slightly simplifies the loop to find the insertion point.
Based on initial patch from Stefano Brivio, including text from the
original patch description too.
The Asus U46E backlight tables have a set of interesting problems:
1. Its ACPI tables do make _OSI ("Windows 2012") checks, so
acpi_osi_is_win8() should return true.
But the tables have 2 sets of _OSI calls, one from the usual global
_INI method setting a global OSYS variable and a second set of _OSI
calls from a MSOS method and the MSOS method is the only one calling
_OSI ("Windows 2012").
The MSOS method only gets called in the following cases:
1. From some Asus specific WMI methods
2. From _DOD, which only runs after acpi_video_get_backlight_type()
has already been called by the i915 driver
3. From other ACPI video bus methods which never run (see below)
4. From some EC query callbacks
So when i915 calls acpi_video_get_backlight_type() MSOS has never run
and acpi_osi_is_win8() returns false, so acpi_video_get_backlight_type()
returns acpi_video as the desired backlight type, which causes
the intel_backlight device to not register.
2. _DOD effectively does this:
Return (Package (0x01)
{
0x0400
})
causing acpi_video_device_in_dod() to return false, which causes
the acpi_video backlight device to not register.
Leaving the user with no backlight device at all. Note that before 6.1.y
the i915 driver would register the intel_backlight device unconditionally
and since that then was the only backlight device userspace would use that.
Add a backlight=native DMI quirk for this special laptop to restore
the old (and working) behavior of the intel_backlight device registering.
Fixes: fb1836c91317 ("ACPI: video: Prefer native over vendor") Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The HP EliteBook 8460p predates Windows 8, so it defaults to using
acpi_video# for backlight control.
Starting with the 6.1.y kernels the native radeon_bl0 backlight is hidden
in this case instead of relying on userspace preferring acpi_video# over
native backlight devices.
It turns out that for the acpi_video# interface to work on
the HP EliteBook 8460p, the brightness needs to be set at least once
through the native interface, which now no longer is done breaking
backlight control.
The native interface however always works without problems, so add
a quirk to use native backlight on the EliteBook 8460p to fix this.
Fixes: fb1836c91317 ("ACPI: video: Prefer native over vendor") Link: https://bugzilla.redhat.com/show_bug.cgi?id=2161428 Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The HP Pavilion g6-1d80nr predates Windows 8, so it defaults to using
acpi_video# for backlight control, but this is non functional on
this model.
Add a DMI quirk to use the native backlight interface which does
work properly.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Stable-dep-of: 9dcb34234b82 ("ACPI: video: Add backlight=native DMI quirk for HP EliteBook 8460p") Signed-off-by: Sasha Levin <sashal@kernel.org>
All nvme transports should be using the same flags for their tagsets,
with the exception for the blocking flag that should only be set for
transports that can block in ->queue_rq.
Add a NVME_F_BLOCKING flag to nvme_ctrl_ops to control the blocking
behavior and lift setting the flags into nvme_alloc_{admin,io}_tag_set.
Allow the transport driver to override the attribute groups for the
control device, so that the PCIe driver doesn't manually have to add a
group after device creation and keep track of it.
The page_pool_release_page was used when freeing rx buffers, and this
function just unmaps the page (if mapped) and does not recycle the page.
So after hundreds of down/up the eth0, the system will out of memory.
For more details, please refer to the following reproduce steps and
bug logs. To solve this issue and refer to the doc of page pool, the
page_pool_put_full_page should be used to replace page_pool_release_page.
Because this API will try to recycle the page if the page refcnt equal to
1. After testing 20000 times, the issue can not be reproduced anymore
(about testing 391 times the issue will occur on i.MX8MN-EVK before).
Reproduce steps:
Create the test script and run the script. The script content is as
follows:
LOOPS=20000
i=1
while [ $i -le $LOOPS ]
do
echo "TINFO:ENET $curface up and down test $i times"
org_macaddr=$(cat /sys/class/net/eth0/address)
ifconfig eth0 down
ifconfig eth0 hw ether $org_macaddr up
i=$(expr $i + 1)
done
sleep 5
if cat /sys/class/net/eth0/operstate | grep 'up';then
echo "TEST PASS"
else
echo "TEST FAIL"
fi
Fixes: 819b93c76f5f ("net: fec: using page pool to manage RX buffers") Signed-off-by: Wei Fang <wei.fang@nxp.com> Reviewed-by: shenwei wang <Shenwei.wang@nxp.com> Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
If net_assign_generic() fails, the current error path in ops_init() tries
to clear the gen pointer slot. Anyway, in such error path, the gen pointer
itself has not been modified yet, and the existing and accessed one is
smaller than the accessed index, causing an out-of-bounds error:
BUG: KASAN: slab-out-of-bounds in ops_init+0x2de/0x320
Write of size 8 at addr ffff888109124978 by task modprobe/1018
Most netlink attributes are parsed and validated from
__nla_validate_parse() or validate_nla()
u16 type = nla_type(nla);
if (type == 0 || type > maxtype) {
/* error or continue */
}
@type is then used as an array index and can be used
as a Spectre v1 gadget.
array_index_nospec() can be used to prevent leaking
content of kernel memory to malicious users.
This should take care of vast majority of netlink uses,
but an audit is needed to take care of others where
validation is not yet centralized in core netlink functions.
iavf_replace_primary_mac() utilizes queue_work() to schedule the
watchdog task but that only ensures that the watchdog task is queued
to run. To make sure the watchdog is executed asap use
mod_delayed_work().
Without this patch it may take up to 2s until the watchdog task gets
executed, which may cause long delays when setting the MAC address.
Fixes: 2b819c0cb66b ("iavf: Add usage of new virtchnl format to set default MAC") Signed-off-by: Stefan Assmann <sassmann@kpanic.de> Reviewed-by: Michal Schmidt <mschmidt@redhat.com> Tested-by: Michal Schmidt <mschmidt@redhat.com> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
We are seeing an issue where setting the MAC address on iavf fails with
EAGAIN after the 2.5s timeout expires in iavf_set_mac().
There is the following deadlock scenario:
iavf_set_mac(), holding rtnl_lock, waits on:
iavf_watchdog_task (within iavf_wq) to send a message to the PF,
and
iavf_adminq_task (within iavf_wq) to receive a response from the PF.
In this adapter state (>=__IAVF_DOWN), these tasks do not need to take
rtnl_lock, but iavf_wq is a global single-threaded workqueue, so they
may get stuck waiting for another adapter's iavf_watchdog_task to run
iavf_init_config_adapter(), which does take rtnl_lock.
The deadlock resolves itself by the timeout in iavf_set_mac(),
which results in EAGAIN returned to userspace.
Let's break the deadlock loop by changing iavf_wq into a per-adapter
workqueue, so that one adapter's tasks are not blocked by another's.
Fixes: 6fb54b1682cb ("iavf: Add waiting for response from PF in set mac") Co-developed-by: Ivan Vecera <ivecera@redhat.com> Signed-off-by: Ivan Vecera <ivecera@redhat.com> Signed-off-by: Michal Schmidt <mschmidt@redhat.com> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
drm_vma_node_allow() and drm_vma_node_revoke() should be called in
balanced pairs. We call drm_vma_node_allow() once per-file everytime a
user calls mmap_offset, but only call drm_vma_node_revoke once per-file
on each mmap_offset. As the mmap_offset is reused by the client, the
per-file vm_count may remain non-zero and the rbtree leaked.
Call drm_vma_node_allow_once() instead to prevent that memory leak.
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Andi Shyti <andi.shyti@linux.intel.com> Signed-off-by: Nirmoy Das <nirmoy.das@intel.com> Fixes: d00cd26a3b7a ("drm/i915/gem: Store mmap_offsets in an rbtree rather than a plain list") Reported-by: Chuansheng Liu <chuansheng.liu@intel.com> Reported-by: Mirsad Todorovac <mirsad.todorovac@alu.unizg.hr> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://lore.kernel.org/r/20230117175236.22317-2-nirmoy.das@intel.com Signed-off-by: Maxime Ripard <maxime@cerno.tech> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently there is no easy way for a drm driver to safely check and allow
drm_vma_offset_node for a drm file just once. Allow drm drivers to call
non-refcounted version of drm_vma_node_allow() so that a driver doesn't
need to keep track of each drm_vma_node_allow() to call subsequent
drm_vma_node_revoke() to prevent memory leak.
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Maxime Ripard <mripard@kernel.org> Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: David Airlie <airlied@gmail.com> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: Andi Shyti <andi.shyti@linux.intel.com> Suggested-by: Chris Wilson <chris.p.wilson@intel.com> Signed-off-by: Nirmoy Das <nirmoy.das@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://lore.kernel.org/r/20230117175236.22317-1-nirmoy.das@intel.com Signed-off-by: Maxime Ripard <maxime@cerno.tech> Signed-off-by: Sasha Levin <sashal@kernel.org>
Ensure that i2c_mark_adapter_suspended() is always balanced by a call to
i2c_mark_adapter_resumed().
dw_i2c_plat_resume() must always be called, so that
i2c_mark_adapter_resumed() is called. This is not compatible with
DPM_FLAG_MAY_SKIP_RESUME, so remove the flag.
Since the controller is always resumed on system resume the
dw_i2c_plat_complete() callback is redundant and has been removed.
The unbalanced suspended flag was introduced by commit 649ad7c38abb
("i2c: designware: Lock the adapter while setting the suspended flag")
Before that commit, the system and runtime PM used the same functions. The
DPM_FLAG_MAY_SKIP_RESUME was used to skip the system resume if the driver
had been in runtime-suspend. If system resume was skipped, the suspended
flag would be cleared by the next runtime resume. The check of the
suspended flag was _after_ the call to pm_runtime_get_sync() in
i2c_dw_xfer(). So either a system resume or a runtime resume would clear
the flag before it was checked.
Having introduced the unbalanced suspended flag with that commit, a further
commit fb49ebadfd4a
("i2c: designware: Use the i2c_mark_adapter_suspended/resumed() helpers")
changed from using a local suspended flag to using the
i2c_mark_adapter_suspended/resumed() functions. These use a flag that is
checked by I2C core code before issuing the transfer to the bus driver, so
there was no opportunity for the bus driver to runtime resume itself before
the flag check.
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com> Fixes: 649ad7c38abb ("i2c: designware: Lock the adapter while setting the suspended flag") Reviewed-by: Hans de Goede <hdegoede@redhat.com> Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In functions i2c_dw_scl_lcnt() and i2c_dw_scl_hcnt() may have overflow
by depending on the values of the given parameters including the ic_clk.
For example in our use case where ic_clk is larger than one million,
multiplication of ic_clk * 4700 will result in 32 bit overflow.
Add cast of u64 to the calculation to avoid multiplication overflow, and
use the corresponding define for divide.
Fixes: 33a0e35a5b24 ("i2c-designware: split of i2c-designware.c into core and bus specific parts") Signed-off-by: Lareine Khawaly <lareine@amazon.com> Signed-off-by: Hanna Hawa <hhhawa@amazon.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Drain requests all go through io_drain_req, which has a quick exit in case
there is nothing pending (ie the drain is not useful). In that case it can
run the issue the request immediately.
However for safety it queues it through task work.
The problem is that in this case the request is run asynchronously, but
the async work has not been prepared through io_req_prep_async.
This has not been a problem up to now, as the task work always would run
before returning to userspace, and so the user would not have a chance to
race with it.
However - with IORING_SETUP_DEFER_TASKRUN - this is no longer the case and
the work might be defered, giving userspace a chance to change data being
referred to in the request.
Instead _always_ prep_async for drain requests, which is simpler anyway
and removes this issue.
A preparation patch, make sure we always hold uring_lock around
io_req_complete_failed(). The only place deviating from the rule
is io_cancel_defer_files(), queue a tw instead.
__io_req_task_work_add() is huge but marked inline, that makes compilers
to generate lots of garbage. Inline the wrapper caller
io_req_task_work_add() instead.
before and after:
text data bss dec hex filename
47347 16248 8 63603 f873 io_uring/io_uring.o
text data bss dec hex filename
45303 16248 8 61559 f077 io_uring/io_uring.o
[Why & How]
Now the vc_start_slot is controlled at drm side. When we
service a long HPD, we still need to run
dm_helpers_dp_mst_write_payload_allocation_table() to update
drm mst_mgr's relevant variable. Otherwise, on the next plug-in,
payload will get assigned with a wrong start slot.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2171 Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Harry Wentland <harry.wentland@amd.com> Fixes: 0828ea1aa8e0 ("drm/display/dp_mst: Move all payload info into the atomic state") Cc: stable@vger.kernel.org # 6.1 Acked-by: Harry Wentland <harry.wentland@amd.com> Reviewed-by: Lyude Paul <lyude@redhat.com> Tested-by: Didier Raboud <odyx@debian.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[Why]
amdgpu expects to update payload table for one stream one time
by calling dm_helpers_dp_mst_write_payload_allocation_table().
Currently, it get modified to try to update HW payload table
at once by referring mst_state.
[How]
This is just a quick workaround. Should find way to remove the
temporary struct dc_dp_mst_stream_allocation_table later if set
struct link_mst_stream_allocatio directly is possible.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2171 Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Harry Wentland <harry.wentland@amd.com> Fixes: 0828ea1aa8e0 ("drm/display/dp_mst: Move all payload info into the atomic state") Cc: stable@vger.kernel.org # 6.1 Acked-by: Harry Wentland <harry.wentland@amd.com> Reviewed-by: Lyude Paul <lyude@redhat.com> Tested-by: Didier Raboud <odyx@debian.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Looks like I made a pretty big mistake here without noticing: it seems when
I moved the assignments of mst_state->pbn_div I completely missed the fact
that the reason for us calling drm_dp_mst_update_slots() earlier was to
account for the fact that we need to call this function using info from the
root MST connector, instead of just trying to do this from each MST
encoder's atomic check function. Otherwise, we end up filling out all of
DC's link information with zeroes.
So, let's restore that and hopefully fix this DSC regression.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2171 Signed-off-by: Lyude Paul <lyude@redhat.com> Signed-off-by: Harry Wentland <harry.wentland@amd.com> Fixes: 0828ea1aa8e0 ("drm/display/dp_mst: Move all payload info into the atomic state") Cc: stable@vger.kernel.org # 6.1 Reviewed-by: Harry Wentland <harry.wentland@amd.com> Tested-by: Didier Raboud <odyx@debian.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Rebase of driver has incorrect unconditional trap enablement
for GFX11 when adding mes queues.
Reported-by: Graham Sider <graham.sider@amd.com> Signed-off-by: Jonathan Kim <jonathan.kim@amd.com> Reviewed-by: Graham Sider <graham.sider@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org # 6.1.x Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[why & how]
We still need to refer to port while removing payload at commit_tail.
we should keep the kref till then to release.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/2171 Signed-off-by: Wayne Lin <Wayne.Lin@amd.com> Signed-off-by: Harry Wentland <harry.wentland@amd.com> Fixes: 0828ea1aa8e0 ("drm/display/dp_mst: Move all payload info into the atomic state") Cc: stable@vger.kernel.org # 6.1 Acked-by: Harry Wentland <harry.wentland@amd.com> Reviewed-by: Lyude Paul <lyude@redhat.com> Tested-by: Didier Raboud <odyx@debian.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
My last commit to fix profile mode displays on AMD platforms caused
an issue on Intel platforms - sorry!
In it I was reading the current functional mode (MMC, PSC, AMT) from
the BIOS but didn't account for the fact that on some of our Intel
platforms I use a different API which returns just the profile and not
the functional mode.
This commit fixes it so that on Intel platforms it knows the functional
mode is always MMC.
I also fixed a potential problem that a platform may try to set the mode
for both MMC and PSC - which was incorrect.
Tested on X1 Carbon 9 (Intel) and Z13 (AMD).
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216963 Fixes: fde5f74ccfc7 ("platform/x86: thinkpad_acpi: Fix profile mode display in AMT mode") Cc: stable@vger.kernel.org Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Mark Pearson <mpearson-lenovo@squebb.ca> Link: https://lore.kernel.org/r/20230124153623.145188-1-mpearson-lenovo@squebb.ca Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The memory for llcc_driv_data is allocated by the LLCC driver. But when
it is passed as the private driver info to the EDAC core, it will get freed
during the qcom_edac driver release. So when the qcom_edac driver gets probed
again, it will try to use the freed data leading to the use-after-free bug.
Hence, do not pass llcc_driv_data as pvt_info but rather reference it
using the platform_data pointer in the qcom_edac driver.
zero_page is a void* pointer but memblock_alloc() returns phys_addr_t type
so this generates a warning while using clang and with -Wint-error enabled
that becomes and error. So let's cast the return of memblock_alloc() to
(void *).
When proxying IPv6 NDP requests, the adverts to the initial multicast
solicits are correct and working. On the other hand, when later a
reachability confirmation is requested (on unicast), no reply is sent.
This causes the neighbor entry expiring on the sending node, which is
mostly a non-issue, as a new multicast request is sent. There are
routers, where the multicast requests are intentionally delayed, and in
these environments the current implementation causes periodic packet
loss for the proxied endpoints.
The root cause is the erroneous decrease of the hop limit, as this
is checked in ndisc.c and no answer is generated when it's 254 instead
of the correct 255.
Cc: stable@vger.kernel.org Fixes: 209ad27be9a8 ("ipv6: decrease hop limit counter in ip6_forward()") Signed-off-by: Gergely Risko <gergely.risko@gmail.com> Tested-by: Gergely Risko <gergely.risko@gmail.com> Reviewed-by: David Ahern <dsahern@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Trip temperatures are read using ACPI methods and stored in the memory
during zone initializtion and when the firmware sends a notification for
change. This trip temperature is returned when the thermal core calls via
callback get_trip_temp().
But it is possible that while updating the memory copy of the trips when
the firmware sends a notification for change, thermal core is reading the
trip temperature via the callback get_trip_temp(). This may return invalid
trip temperature.
To address this add a mutex to protect the invalid temperature reads in
the callback get_trip_temp() and int340x_thermal_read_trips().
Fixes: c6f47ab05910 ("Thermal/int340x: Add common thermal zone handler") Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: 5.0+ <stable@vger.kernel.org> # 5.0+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since commit 80b6093b55e3 ("kbuild: add -Wundef to KBUILD_CPPFLAGS
for W=1 builds"), building with W=1 detects misuse of #if.
$ make W=1 ARCH=riscv CROSS_COMPILE=riscv64-linux-gnu- arch/riscv/kernel/
[snip]
AS arch/riscv/kernel/head.o
arch/riscv/kernel/head.S:329:5: warning: "CONFIG_RISCV_BOOT_SPINWAIT" is not defined, evaluates to 0 [-Wundef]
329 | #if CONFIG_RISCV_BOOT_SPINWAIT
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
CONFIG_RISCV_BOOT_SPINWAIT is a bool option. #ifdef should be used.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Fixes: e57b14b4666e ("RISC-V: Move spinwait booting method to its own config") Link: https://lore.kernel.org/r/20230106161213.2374093-1-masahiroy@kernel.org Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a lock inversion and rwsem read-lock recursion in the devfreq
target callback which can lead to deadlocks.
Specifically, ufshcd_devfreq_scale() already holds a clk_scaling_lock
read lock when toggling the write booster, which involves taking the
dev_cmd mutex before taking another clk_scaling_lock read lock.
This can lead to a deadlock if another thread:
1) tries to acquire the dev_cmd and clk_scaling locks in the correct
order, or
2) takes a clk_scaling write lock before the attempt to take the
clk_scaling read lock a second time.
Fix this by dropping the clk_scaling_lock before toggling the write booster
as was done before commit 88cfe29932e9 ("scsi: ufs: Protect some contexts
from unexpected clock scaling").
While the devfreq callbacks are already serialised, add a second
serialising mutex to handle the unlikely case where a callback triggered
through the devfreq sysfs interface is racing with a request to disable
clock scaling through the UFS controller 'clkscale_enable' sysfs
attribute. This could otherwise lead to the write booster being left
disabled after having disabled clock scaling.
Also take the new mutex in ufshcd_clk_scaling_allow() to make sure that any
pending write booster update has completed on return.
Note that this currently only affects Qualcomm platforms since commit bcd3c82c371a ("scsi: ufs: core: Allow host driver to disable wb toggling
during clock scaling").
The lock inversion (i.e. 1 above) was reported by lockdep as:
======================================================
WARNING: possible circular locking dependency detected
6.1.0-next-20221216 #211 Not tainted
------------------------------------------------------
kworker/u16:2/71 is trying to acquire lock: ffff076280ba98a0 (&hba->dev_cmd.lock){+.+.}-{3:3}, at: ufshcd_query_flag+0x50/0x1c0
but task is already holding lock: ffff076280ba9cf0 (&hba->clk_scaling_lock){++++}-{3:3}, at: ufshcd_devfreq_scale+0x2b8/0x380
which lock already depends on the new lock.
[ +0.011606]
the existing dependency chain (in reverse order) is:
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&hba->clk_scaling_lock);
lock(&hba->dev_cmd.lock);
lock(&hba->clk_scaling_lock);
lock(&hba->dev_cmd.lock);
*** DEADLOCK ***
Fixes: 88cfe29932e9 ("scsi: ufs: Protect some contexts from unexpected clock scaling") Cc: stable@vger.kernel.org # 5.12 Cc: Can Guo <quic_cang@quicinc.com> Tested-by: Andrew Halaney <ahalaney@redhat.com> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Reviewed-by: Bart Van Assche <bvanassche@acm.org> Link: https://lore.kernel.org/r/20230116161201.16923-1-johan+linaro@kernel.org Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To save the vgic LPI pending state with GICv4.1, the VPEs must all be
unmapped from the ITSs so that the sGIC caches can be flushed.
The opposite is done once the state is saved.
This is all done by using the activate/deactivate irqdomain callbacks
directly from the vgic code. Crutially, this is done without holding
the irqdesc lock for the interrupts that represent the VPE. And these
callbacks are changing the state of the irqdesc. What could possibly
go wrong?
If a doorbell fires while we are messing with the irqdesc state,
it will acquire the lock and change the interrupt state concurrently.
Since we don't hole the lock, curruption occurs in on the interrupt
state. Oh well.
While acquiring the lock would fix this (and this was Shanker's
initial approach), this is still a layering violation we could do
without. A better approach is actually to free the VPE interrupt,
do what we have to do, and re-request it.
It is more work, but this usually happens only once in the lifetime
of the VM and we don't really care about this sort of overhead.
Fixes: 4f13f163c444 ("KVM: arm64: GICv4.1: Try to save VLPI state in save_pending_tables") Reported-by: Shanker Donthineni <sdonthineni@nvidia.com> Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20230118022348.4137094-1-sdonthineni@nvidia.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When serializing and deserializing kvm_sregs, attributes of the segment
descriptors are stored by user space. For unusable segments,
vmx_segment_access_rights skips all attributes and sets them to 0.
This means we zero out the DPL (Descriptor Privilege Level) for unusable
entries.
Unusable segments are - contrary to their name - usable in 64bit mode and
are used by guests to for example create a linear map through the
NULL selector.
VMENTER checks if SS.DPL is correct depending on the CS segment type.
For types 9 (Execute Only) and 11 (Execute Read), CS.DPL must be equal to
SS.DPL [1].
We have seen real world guests setting CS to a usable segment with DPL=3
and SS to an unusable segment with DPL=3. Once we go through an sregs
get/set cycle, SS.DPL turns to 0. This causes the virtual machine to crash
reproducibly.
This commit changes the attribute logic to always preserve attributes for
unusable segments. According to [2] SS.DPL is always saved on VM exits,
regardless of the unusable bit so user space applications should have saved
the information on serialization correctly.
[3] specifies that besides SS.DPL the rest of the attributes of the
descriptors are undefined after VM entry if unusable bit is set. So, there
should be no harm in setting them all to the previous state.
If we're using ring provided buffers with multishot receive, and we end
up doing an io-wq based issue at some points that also needs to select
a buffer, we'll lose the initially assigned buffer group as
io_ring_buffer_select() correctly clears the buffer group list as the
issue isn't serialized by the ctx uring_lock. This is fine for normal
receives as the request puts the buffer and finishes, but for multishot,
we will re-arm and do further receives. On the next trigger for this
multishot receive, the receive will try and pick from a buffer group
whose value is the same as the buffer ID of the las receive. That is
obviously incorrect, and will result in a premature -ENOUFS error for
the receive even if we had available buffers in the correct group.
Cache the buffer group value at prep time, so we can restore it for
future receives. This only needs doing for the above mentioned case, but
just do it by default to keep it easier to read.
If st_uid/st_gid doesn't have a mapping in the mounter's user_ns, then
copy-up should fail, just like it would fail if the mounter task was doing
the copy using "cp -a".
There's a corner case where the "cp -a" would succeed but copy up fail: if
there's a mapping of the invalid uid/gid (65534 by default) in the user
namespace. This is because stat(2) will return this value if the mapping
doesn't exist in the current user_ns and "cp -a" will in turn be able to
create a file with this uid/gid.
This behavior would be inconsistent with POSIX ACL's, which return -1 for
invalid uid/gid which result in a failed copy.
For consistency and simplicity fail the copy of the st_uid/st_gid are
invalid.
Stream protocol length will never be larger than 16KB until session setup.
After session setup, the size of requests will not be larger than
16KB + SMB2 MAX WRITE size. This patch limits these invalidly oversized
requests and closes the connection immediately.
Fixes: 34342fc8255d ("cifsd: add server handler for central processing and tranport layers") Cc: stable@vger.kernel.org Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-18259 Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When user switch samba to ksmbd, The following message flood is coming
when accessing files. Samba seems to changs dos attribute version to v5.
This patch downgrade ndr version error message to debug.
$ dmesg
...
[68971.766914] ksmbd: v5 version is not supported
[68971.779808] ksmbd: v5 version is not supported
[68971.871544] ksmbd: v5 version is not supported
[68971.910135] ksmbd: v5 version is not supported
...
Cc: stable@vger.kernel.org Fixes: 9445e1eebdb7 ("cifsd: add server-side procedures for SMB3") Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If ksmbd.mountd is configured to assign unknown users to the guest account
("map to guest = bad user" in the config), ksmbd signs the response.
This is wrong according to MS-SMB2 3.3.5.5.3:
12. If the SMB2_SESSION_FLAG_IS_GUEST bit is not set in the SessionFlags
field, and Session.IsAnonymous is FALSE, the server MUST sign the
final session setup response before sending it to the client, as
follows:
[...]
This fixes libsmb2 based applications failing to establish a session
("Wrong signature in received").
Fixes: 9445e1eebdb7 ("cifsd: add server-side procedures for SMB3") Cc: stable@vger.kernel.org Signed-off-by: Marios Makassikis <mmakassikis@freebox.fr> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add max connections parameter to limit number of maximum simultaneous
connections.
Fixes: 34342fc8255d ("cifsd: add server handler for central processing and tranport layers") Cc: stable@vger.kernel.org Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In smbd_destroy(), clear the server->smbd_conn pointer after freeing the
smbd_connection struct that it points to so that reconnection doesn't get
confused.
Fixes: 2cb924bb40fd ("CIFS: SMBD: Implement function to destroy a SMB Direct connection") Cc: stable@vger.kernel.org Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Acked-by: Tom Talpey <tom@talpey.com> Signed-off-by: David Howells <dhowells@redhat.com> Cc: Long Li <longli@microsoft.com> Cc: Pavel Shilovsky <piastryyy@gmail.com> Cc: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The instructions for the ftrace-bisect.sh script, which is used to find
what function is being traced that is causing a kernel crash, and possibly
a triple fault reboot, uses the old method. In 5.1, a new feature was
added that let the user write in the index into available_filter_functions
that maps to the function a user wants to set in set_ftrace_filter (or
set_ftrace_notrace). This takes O(1) to set, as suppose to writing a
function name, which takes O(n) (where n is the number of functions in
available_filter_functions).
The ftrace-bisect.sh requires setting half of the functions in
available_filter_functions, which is O(n^2) using the name method to enable
and can take several minutes to complete. The number method is O(n) which
takes less than a second to complete. Using the number method for any
kernel 5.1 and after is the proper way to do the bisect.
Update the usage to reflect the new change, as well as using the
/sys/kernel/tracing path instead of the obsolete debugfs path.
Link: https://lkml.kernel.org/r/20230123112252.022003dd@gandalf.local.home Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Acked-by: Mark Rutland <mark.rutland@arm.com> Fixes: a75673e3593b4 ("ftrace: Allow enabling of filters via index of available_filter_functions") Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Function 'create_hist_field' is called recursively at
trace_events_hist.c:1954 and can return NULL-value that's why we have
to check it to avoid null pointer dereference.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Link: https://lkml.kernel.org/r/20230111120409.4111-1-n.petrova@fintech.ru Cc: stable@vger.kernel.org Fixes: 6ac0075044a9 ("tracing: Add variable support to hist triggers") Signed-off-by: Natalia Petrova <n.petrova@fintech.ru> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently trace_printk() can be used as soon as early_trace_init() is
called from start_kernel(). But if a crash happens, and
"ftrace_dump_on_oops" is set on the kernel command line, all you get will
be:
[ 0.456075] <idle>-0 0dN.2. 347519us : Unknown type 6
[ 0.456075] <idle>-0 0dN.2. 353141us : Unknown type 6
[ 0.456075] <idle>-0 0dN.2. 358684us : Unknown type 6
This is because the trace_printk() event (type 6) hasn't been registered
yet. That gets done via an early_initcall(), which may be early, but not
early enough.
Instead of registering the trace_printk() event (and other ftrace events,
which are not trace events) via an early_initcall(), have them registered at
the same time that trace_printk() can be used. This way, if there is a
crash before early_initcall(), then the trace_printk()s will actually be
useful.
Link: https://lkml.kernel.org/r/20230104161412.019f6c55@gandalf.local.home Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Fixes: 546b9546f5307 ("tracing: Split tracing initialization into two for early initialization") Reported-by: "Joel Fernandes (Google)" <joel@joelfernandes.org> Tested-by: Joel Fernandes (Google) <joel@joelfernandes.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Setting filters on an ftrace ops results in some memory being allocated
for the filter hashes, which must be freed before the ops can be freed.
This can be done by removing every individual element of the hash by
calling ftrace_set_filter_ip() or ftrace_set_filter_ips() with `remove`
set, but this is somewhat error prone as it's easy to forget to remove
an element.
Make it easier to clean this up by exporting ftrace_free_filter(), which
can be used to clean up all of the filter hashes after an ftrace_ops has
been unregistered.
Using this, fix the ftrace-direct* samples to free hashes prior to being
unloaded. All other code either removes individual filters explicitly or
is built-in and already calls ftrace_free_filter().
Link: https://lkml.kernel.org/r/20230103124912.2948963-3-mark.rutland@arm.com Cc: stable@vger.kernel.org Cc: Florent Revest <revest@chromium.org> Cc: Masami Hiramatsu <mhiramat@kernel.org> Fixes: 032508183d58 ("ftrace/samples: Add module to test multi direct modify interface") Fixes: 7b3e3a4a7378 ("ftrace/samples: Add multi direct interface test module") Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During a system boot, it can happen that the kernel receives a burst of
requests to insert the same module but loading it eventually fails
during its init call. For instance, udev can make a request to insert
a frequency module for each individual CPU when another frequency module
is already loaded which causes the init function of the new module to
return an error.
Since commit 44dfd5c6e62d ("kernel/module.c: Only return -EEXIST for
modules that have finished loading"), the kernel waits for modules in
MODULE_STATE_GOING state to finish unloading before making another
attempt to load the same module.
This creates unnecessary work in the described scenario and delays the
boot. In the worst case, it can prevent udev from loading drivers for
other devices and might cause timeouts of services waiting on them and
subsequently a failed boot.
This patch attempts a different solution for the problem 44dfd5c6e62d
was trying to solve. Rather than waiting for the unloading to complete,
it returns a different error code (-EBUSY) for modules in the GOING
state. This should avoid the error situation that was described in 44dfd5c6e62d (user space attempting to load a dependent module because
the -EEXIST error code would suggest to user space that the first module
had been loaded successfully), while avoiding the delay situation too.
This has been tested on linux-next since December 2022 and passes
all kmod selftests except test 0009 with module compression enabled
but it has been confirmed that this issue has existed and has gone
unnoticed since prior to this commit and can also be reproduced without
module compression with a simple usleep(5000000) on tools/modprobe.c [0].
These failures are caused by hitting the kernel mod_concurrent_max and can
happen either due to a self inflicted kernel module auto-loead DoS somehow
or on a system with large CPU count and each CPU count incorrectly triggering
many module auto-loads. Both of those issues need to be fixed in-kernel.
Fixes: 44dfd5c6e62d ("kernel/module.c: Only return -EEXIST for modules that have finished loading") Co-developed-by: Martin Wilck <mwilck@suse.com> Signed-off-by: Martin Wilck <mwilck@suse.com> Signed-off-by: Petr Pavlu <petr.pavlu@suse.com> Cc: stable@vger.kernel.org Reviewed-by: Petr Mladek <pmladek@suse.com>
[mcgrof: enhance commit log with testing and kmod test result interpretation ] Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
nfsd_file_cache_purge is called when the server is shutting down, in
which case, tearing things down is generally fine, but it also gets
called when the exports cache is flushed.
Instead of walking the cache and freeing everything unconditionally,
handle it the same as when we have a notification of conflicting access.
Currently it is possible that the final put of a KVM reference comes from
vfio during its device close operation. This occurs while the vfio group
lock is held; however, if the vfio device is still in the kvm device list,
then the following call chain could result in a deadlock:
VFIO holds group->group_lock/group_rwsem
-> kvm_put_kvm
-> kvm_destroy_vm
-> kvm_destroy_devices
-> kvm_vfio_destroy
-> kvm_vfio_file_set_kvm
-> vfio_file_set_kvm
-> try to hold group->group_lock/group_rwsem
The key function is the kvm_destroy_devices() which triggers destroy cb
of kvm_device_ops. It calls back to vfio and try to hold group_lock. So
if this path doesn't call back to vfio, this dead lock would be fixed.
Actually, there is a way for it. KVM provides another point to free the
kvm-vfio device which is the point when the device file descriptor is
closed. This can be achieved by providing the release cb instead of the
destroy cb. Also rename kvm_vfio_destroy() to be kvm_vfio_release().
/*
* Destroy is responsible for freeing dev.
*
* Destroy may be called before or after destructors are called
* on emulated I/O regions, depending on whether a reference is
* held by a vcpu or other kvm component that gets destroyed
* after the emulated I/O.
*/
void (*destroy)(struct kvm_device *dev);
/*
* Release is an alternative method to free the device. It is
* called when the device file descriptor is closed. Once
* release is called, the destroy method will not be called
* anymore as the device is removed from the device list of
* the VM. kvm->lock is held.
*/
void (*release)(struct kvm_device *dev);
Fixes: 39a6b4e8fcdb ("vfio: remove VFIO_GROUP_NOTIFY_SET_KVM") Reported-by: Alex Williamson <alex.williamson@redhat.com> Suggested-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com> Link: https://lore.kernel.org/r/20230114000351.115444-1-mjrosato@linux.ibm.com Link: https://lore.kernel.org/r/20230120150528.471752-1-yi.l.liu@intel.com
[aw: update comment as well, s/destroy/release/] Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The 'h' is a pointer to struct ctlr_info, so it's just 4 or 8 bytes, while
the structure itself is much bigger.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: ef421ae82373 ("hpsa: add driver for HP Smart Array controllers.") Link: https://lore.kernel.org/r/20230118031255.GE15213@altlinux.org Signed-off-by: Alexey V. Vissarionov <gremlin@altlinux.org> Acked-by: Don Brace <don.brace@microchip.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since commit cbf7827bc5dc ("iommu/s390: Fix potential s390_domain
aperture shrinking") the s390 IOMMU driver uses reserved regions for the
system provided DMA ranges of PCI devices. Previously it reduced the
size of the IOMMU aperture and checked it on each mapping operation.
On current machines the system denies use of DMA addresses below 2^32 for
all PCI devices.
Usually mapping IOVAs in a reserved regions is harmless until a DMA
actually tries to utilize the mapping. However on s390 there is
a virtual PCI device called ISM which is implemented in firmware and
used for cross LPAR communication. Unlike real PCI devices this device
does not use the hardware IOMMU but inspects IOMMU translation tables
directly on IOTLB flush (s390 RPCIT instruction). If it detects IOVA
mappings outside the allowed ranges it goes into an error state. This
error state then causes the device to be unavailable to the KVM guest.
Analysing this we found that vfio_test_domain_fgsp() maps 2 pages at DMA
address 0 irrespective of the IOMMUs reserved regions. Even if usually
harmless this seems wrong in the general case so instead go through the
freshly updated IOVA list and try to find a range that isn't reserved,
and fits 2 pages, is PAGE_SIZE * 2 aligned. If found use that for
testing for fine grained super pages.
Fixes: 828c9fc230e0 ("vfio/type1: Check reserved region conflict and update iova list") Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com> Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/20230110164427.4051938-2-schnelle@linux.ibm.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When evaluating the CPU candidates in the perf domain (pd) containing
the previously used CPU (prev_cpu), find_energy_efficient_cpu()
evaluates the energy of the pd:
- without the task (base_energy)
- with the task placed on prev_cpu (if the task fits)
- with the task placed on the CPU with the highest spare capacity,
prev_cpu being excluded from this set
If prev_cpu is already the CPU with the highest spare capacity,
max_spare_cap_cpu will be the CPU with the second highest spare
capacity.
On an Arm64 Juno-r2, with a workload of 10 tasks at a 10% duty cycle,
when prev_cpu and max_spare_cap_cpu are both valid candidates,
prev_spare_cap > max_spare_cap at ~82%.
Thus the energy of the pd when placing the task on max_spare_cap_cpu
is computed with no possible positive outcome 82% most of the time.
Do not consider max_spare_cap_cpu as a valid candidate if
prev_spare_cap > max_spare_cap.
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com> Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org> Link: https://lore.kernel.org/r/20221006081052.3862167-2-pierre.gondois@arm.com
Stable-dep-of: e26fd28db828 ("sched/uclamp: Fix a uninitialized variable warnings") Signed-off-by: Sasha Levin <sashal@kernel.org>
When a running wake_tx_queue() call is aborted due to a hw queue stop
the corresponding iTXQ is not always correctly marked for resumption:
wake_tx_push_queue() can stops the queue run without setting
@IEEE80211_TXQ_STOP_NETIF_TX.
Without the @IEEE80211_TXQ_STOP_NETIF_TX flag __ieee80211_wake_txqs()
will not schedule a new queue run and remaining frames in the queue get
stuck till another frame is queued to it.
Fix the issue for all drivers - also the ones with custom wake_tx_queue
callbacks - by moving the logic into ieee80211_tx_dequeue() and drop the
redundant @txqs_stopped.
@IEEE80211_TXQ_STOP_NETIF_TX is also renamed to @IEEE80211_TXQ_DIRTY to
better describe the flag.
Fixes: c850e31f79f0 ("wifi: mac80211: add internal handler for wake_tx_queue") Signed-off-by: Alexander Wetzel <alexander@wetzel-home.de> Link: https://lore.kernel.org/r/20221230121850.218810-1-alexander@wetzel-home.de Cc: stable@vger.kernel.org Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
IORING_SETUP_R_DISABLED rings don't have the submitter task set, so
it's not always safe to use ->submitter_task. Disallow posting msg_ring
messaged to disabled rings. Also add task NULL check for loosy sync
around testing for IORING_SETUP_R_DISABLED.
Cc: stable@vger.kernel.org Fixes: 6d043ee1164ca ("io_uring: do msg_ring in target task via tw") Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stop considering VBT's static DRRS support when deciding whether
to use alternate fixed modes or not. It looks like Windows more
or less just uses that to decide whether to automagically switch
refresh rates on AC<->battery changes, or perhaps whether to
even expose a control for that in some UI thing. Either way it
seems happy to always use all EDID modes, and I guess the
DRRS/VRR stuff more or less adjusts how said modes get
actually used.
Let's do the same and just accept all the suitable looking
modes from EDID, whether we have DRRS or VRR.
Apparently some panels declare multiple modes with random
sync polarities. Seems a bit weird, but looks like Windows/GOP
doesn't care, so let follow suit and accept alternate fixed
modes regardless of their sync polarities.
v2: Don't pollute the DRM_ namespace with a define (Jani)
Comparing current_work() against efi_rts_work.work is sufficient to
decide whether current is currently running EFI runtime services code at
any level in its call stack.
However, there are other potential users of the EFI runtime stack, such
as the ACPI subsystem, which may invoke efi_call_virt_pointer()
directly, and so any sync exceptions occurring in firmware during those
calls are currently misidentified.
So instead, let's check whether the stashed value of the thread stack
pointer points into current's thread stack. This can only be the case if
current was interrupted while running EFI runtime code. Note that this
implies that we should clear the stashed value after switching back, to
avoid false positives.
Unlike x86, which has machinery to deal with page faults that occur
during the execution of EFI runtime services, arm64 has nothing like
that, and a synchronous exception raised by firmware code brings down
the whole system.
With more EFI based systems appearing that were not built to run Linux
(such as the Windows-on-ARM laptops based on Qualcomm SOCs), as well as
the introduction of PRM (platform specific firmware routines that are
callable just like EFI runtime services), we are more likely to run into
issues of this sort, and it is much more likely that we can identify and
work around such issues if they don't bring down the system entirely.
Since we already use a EFI runtime services call wrapper in assembler,
we can quite easily add some code that captures the execution state at
the point where the call is made, allowing us to revert to this state
and proceed execution if the call triggered a synchronous exception.
Given that the kernel and the firmware don't share any data structures
that could end up in an indeterminate state, we can happily continue
running, as long as we mark the EFI runtime services as unavailable from
that point on.
The commit f4e76fcfba27 ("btrfs: zoned: disable metadata overcommit for
zoned") disabled the metadata over-commit to track active zones properly.
However, it also introduced a heavy overhead by allocating new metadata
block groups and/or flushing dirty buffers to release the space
reservations. Specifically, a workload (write only without any sync
operations) worsen its performance from 343.77 MB/sec (v5.19) to 182.89
MB/sec (v6.0).
The performance is still bad on current misc-next which is 187.95 MB/sec.
And, with this patch applied, it improves back to 326.70 MB/sec (+73.82%).
This patch introduces a new fs_info->flag BTRFS_FS_NO_OVERCOMMIT to
indicate it needs to disable the metadata over-commit. The flag is enabled
when a device with max active zones limit is loaded into a file-system.
Fixes: f4e76fcfba27 ("btrfs: zoned: disable metadata overcommit for zoned") CC: stable@vger.kernel.org # 6.0+ Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The memcpy() of the data following a coreboot_table_entry couldn't
be evaluated by the compiler under CONFIG_FORTIFY_SOURCE. To make it
easier to reason about, add an explicit flexible array member to struct
coreboot_device so the entire entry can be copied at once. Additionally,
validate the sizes before copying. Avoids this run-time false positive
warning:
memcpy: detected field-spanning write (size 168) of single field "&device->entry" at drivers/firmware/google/coreboot_table.c:103 (size 8)
This driver uses MSR functions that aren't implemented under UML.
Avoid building it to prevent tripping up allyesconfig.
e.g.
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x3a3): undefined reference to `__tracepoint_read_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x3d2): undefined reference to `__tracepoint_write_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x457): undefined reference to `__tracepoint_write_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x481): undefined reference to `do_trace_write_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x4d5): undefined reference to `do_trace_write_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x4f5): undefined reference to `do_trace_read_msr'
/usr/lib/gcc/x86_64-pc-linux-gnu/12/../../../../x86_64-pc-linux-gnu/bin/ld: pata_cs5535.c:(.text+0x51c): undefined reference to `do_trace_write_msr'
Signed-off-by: Peter Foley <pefoley2@pefoley.com> Reviewed-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Damien Le Moal <damien.lemoal@opensource.wdc.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
On the x86-64 architecture even a failing cmpxchg grants exclusive
access to the cacheline, making it preferable to retry the failed op
immediately instead of stalling with the pause instruction.
To illustrate the impact, below are benchmark results obtained by
running various will-it-scale tests on top of the 6.2-rc3 kernel and
Cascade Lake (2 sockets * 24 cores * 2 threads) CPU.
All results in ops/s. Note there is some variance in re-runs, but the
code is consistently faster when contention is present.
That is, fixing the problems in access itself *reduces* scalability
after the cacheline ping-pong only happens in lockref with the pause
instruction.
Note that fstat and access benchmarks are not currently integrated into
will-it-scale, but interested parties can find them in pull requests to
said project.
Code at hand has a rather tortured history. First modification showed
up in commit ea82cc601b2b ("lockref: Relax in cmpxchg loop"), written
with Itanium in mind. Later it got patched up to use an arch-dependent
macro to stop doing it on s390 where it caused a significant regression.
Said macro had undergone revisions and was ultimately eliminated later,
going back to cpu_relax.
While I intended to only remove cpu_relax for x86-64, I got the
following comment from Linus:
I would actually prefer just removing it entirely and see if
somebody else hollers. You have the numbers to prove it hurts on
real hardware, and I don't think we have any numbers to the
contrary.
So I think it's better to trust the numbers and remove it as a
failure, than say "let's just remove it on x86-64 and leave
everybody else with the potentially broken code"
Additionally, Will Deacon (maintainer of the arm64 port, one of the
architectures previously benchmarked):
So, from the arm64 side of the fence, I'm perfectly happy just
removing the cpu_relax() calls from lockref.
As such, come back full circle in history and whack it altogether.
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com> Link: https://lore.kernel.org/all/CAGudoHHx0Nqg6DE70zAVA75eV-HXfWyhVMWZ-aSeOofkA_=WdA@mail.gmail.com/ Acked-by: Tony Luck <tony.luck@intel.com> # ia64 Acked-by: Nicholas Piggin <npiggin@gmail.com> # powerpc Acked-by: Will Deacon <will@kernel.org> # arm64 Acked-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some laptops have a fan device listed in their ACPI tables but do not
actually contain a fan.
Introduce a quirk that can be used to override the fan detection logic.
This was observed with a ASUS VivoBook E410MA running firmware
E410MAB.304.
The Microsoft Devkit 2023 is a an ARM64 based machine featuring a
Realtek 8153 USB3.0-to-GBit Ethernet adapter. As in their other
machines, Microsoft uses a custom USB device ID.
Add the respective ID values to the driver. This makes Ethernet work on
the MS Devkit device. The chip has been visually confirmed to be a
RTL8153.
Currently the driver sets the port invalid if one phy in the port is not
enabled, which may cause issues in expander situation. In directly attached
situation, if phy up doesn't occur in time when refreshing port id, the
port is incorrectly set to invalid which will also cause disk lost.
Therefore set a port invalid only if there are no devices attached to the
port.
Currently clear task set is used to abort all commands remaining in the
disk when the SAS disk is discovered, and if the disk is discovered by two
initiators, other I_T nexuses are also affected. So use abort task set
instead and take effect only on the specified I_T nexus.
The two debug messages in spidev_open() dereference spidev->spi without
taking the lock and without checking if it's not null. This can lead to
a crash. Drop the messages as they're not needed - the user-space will
get informed about ENOMEM with the syscall return value.
There will be data corruption on vram allocated by svm
if the initialization is not complete and application is
writting on the memory. Adding sync to wait for the
initialization completion is to resolve this issue.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com> Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Like the Asus Expertbook B2502CBA and various Asus Vivobook laptops,
the Asus Expertbook B2402CBA has an ACPI DSDT table that describes IRQ 1
as ActiveLow while the kernel overrides it to Edge_High. This prevents the
keyboard from working. To fix this issue, add this laptop to the
skip_override_table so that the kernel does not override IRQ 1.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216864 Tested-by: zelenat <zelenat@gmail.com> Signed-off-by: Tamim Khan <tamim@fusetak.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The fsl-asoc-card AC'97 support currently tries to route to Playback and
Capture widgets provided by the AC'97 CODEC. This doesn't work since the
generic AC'97 driver registers with an "AC97" at the front of the stream
and hence widget names, update to reflect reality. It's not clear to me
if or how this ever worked.
The SSI driver calls the AC'97 playback and transmit streams "AC97 Playback"
and "AC97 Capture" respectively. This is the same name used by the generic
AC'97 CODEC driver in ASoC, creating confusion for the Freescale ASoC card
when it attempts to use these widgets in routing. Add a "CPU" in the name
like the regular DAIs registered by the driver to disambiguate.
Use NULL for NULL pointer to fix the following sparse warning:
drivers/cpufreq/armada-37xx-cpufreq.c:448:32: sparse: warning: Using plain integer as NULL pointer
When building on ARM in thumb mode with gcc-11.3 at -O2 or -O3,
nolibc-test segfaults during the select() tests. It turns out that at
this level, gcc recognizes an opportunity for using memset() to zero
the fd_set, but it miscompiles it because it also recognizes a memset
pattern as well, and decides to call memset() from the memset() code:
projects / kernel.git / log