RDMA is not supported in ice on a PF that has been added to a bonded
interface. To enforce this, when an interface enters a bond, we unplug
the auxiliary device that supports RDMA functionality. This unplug
currently happens in the context of handling the netdev bonding event.
This event is sent to the ice driver under RTNL context. This is causing
a deadlock where the RDMA driver is waiting for the RTNL lock to complete
the removal.
Defer the unplugging/re-plugging of the auxiliary device to the service
task so that it is not performed under the RTNL lock context.
Cc: stable@vger.kernel.org # 6.1.x Reported-by: Jaroslav Pulchart <jaroslav.pulchart@gooddata.com> Link: https://lore.kernel.org/netdev/CAK8fFZ6A_Gphw_3-QMGKEFQk=sfCw1Qmq0TVZK3rtAi7vb621A@mail.gmail.com/ Fixes: a276ea96090c ("ice: Fix race condition during interface enslave") Fixes: 91e1f22ab3c9 ("RDMA/irdma: Report the correct link speed") Signed-off-by: Dave Ertman <david.m.ertman@intel.com> Tested-by: Arpana Arland <arpanax.arland@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Link: https://lore.kernel.org/r/20230310194833.3074601-1-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When buffered write fails to copy data into underlying page cache page,
ocfs2_write_end_nolock() just zeroes out and dirties the page. This can
leave dirty page beyond EOF and if page writeback tries to write this page
before write succeeds and expands i_size, page gets into inconsistent
state where page dirty bit is clear but buffer dirty bits stay set
resulting in page data never getting written and so data copied to the
page is lost. Fix the problem by invalidating page beyond EOF after
failed write.
Link: https://lkml.kernel.org/r/20230302153843.18499-1-jack@suse.cz Fixes: 31a7abf11e72 ("fs: Don't invalidate page buffers in block_write_full_page()") Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> Cc: Mark Fasheh <mark@fasheh.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Junxiao Bi <junxiao.bi@oracle.com> Cc: Changwei Ge <gechangwei@live.cn> Cc: Gang He <ghe@suse.com> Cc: Jun Piao <piaojun@huawei.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
BUG: KASAN: use-after-free in lookup_rec
Read of size 8 at addr ffff000199270ff0 by task modprobe
CPU: 2 Comm: modprobe
Call trace:
kasan_report
__asan_load8
lookup_rec
ftrace_location
arch_check_ftrace_location
check_kprobe_address_safe
register_kprobe
When checking pg->records[pg->index - 1].ip in lookup_rec(), it can get a
pg which is newly added to ftrace_pages_start in ftrace_process_locs().
Before the first pg->index++, index is 0 and accessing pg->records[-1].ip
will cause this problem.
Don't check the ip when pg->index is 0.
Link: https://lore.kernel.org/linux-trace-kernel/20230309080230.36064-1-chenzhongjin@huawei.com Cc: stable@vger.kernel.org Fixes: a1fa044f2b61 ("ftrace: Speed up search by skipping pages by address") Suggested-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Christoph reports a lockdep splat in the mptcp_subflow_create_socket()
error path, when such function is invoked by
mptcp_pm_nl_create_listen_socket().
Such code path acquires two separates, nested socket lock, with the
internal lock operation lacking the "nested" annotation. Adding that
in sock_release() for mptcp's sake only could be confusing.
Instead just add a new lockclass to the in-kernel msk socket,
re-initializing the lockdep infra after the socket creation.
Fixes: 8f1319551a64 ("mptcp: fix locking for in-kernel listener creation") Cc: stable@vger.kernel.org Reported-by: Christoph Paasch <cpaasch@apple.com> Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/354 Signed-off-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net> Tested-by: Christoph Paasch <cpaasch@apple.com> Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add __ro_after_init labels for the variables tcp_prot_override and
tcpv6_prot_override, just like other variables adjacent to them, to
indicate that they are initialised from the init hooks and no writes
occur afterwards.
Christoph reported a possible deadlock while the TCP stack
destroys an unaccepted subflow due to an incoming reset: the
MPTCP socket error path tries to acquire the msk-level socket
lock while TCP still owns the listener socket accept queue
spinlock, and the reverse dependency already exists in the
TCP stack.
Note that the above is actually a lockdep false positive, as
the chain involves two separate sockets. A different per-socket
lockdep key will address the issue, but such a change will be
quite invasive.
Instead, we can simply stop earlier the socket error handling
for orphaned or unaccepted subflows, breaking the critical
lockdep chain. Error handling in such a scenario is a no-op.
Reported-and-tested-by: Christoph Paasch <cpaasch@apple.com> Fixes: d9b41d28dae4 ("mptcp: deliver ssk errors to msk") Cc: stable@vger.kernel.org Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/355 Signed-off-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net> Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Always setup overdrive tables after resume. Preserve only some
user-defined settings in user_overdrive_table if they're set.
Copy restored user_overdrive_table into od_table to get correct
values.
On cold boot, BTC was triggered and GfxVfCurve was calibrated. We
got VfCurve settings (a). On resuming back, BTC will be triggered
again and GfxVfCurve will be recalibrated. VfCurve settings (b)
got may be different from those of cold boot. So if we reuse
those VfCurve settings (a) got on cold boot on suspend, we can
run into discrepencies.
Check kfd->init_complete in kgd2kfd_iommu_resume, consistent with other
kgd2kfd calls. This should fix IOMMU errors on resume from suspend when
KFD IOMMU initialization failed.
Users reported oopses on list corruptions when using i915 perf with a
number of concurrently running graphics applications. Root cause analysis
pointed at an issue in barrier processing code -- a race among perf open /
close replacing active barriers with perf requests on kernel context and
concurrent barrier preallocate / acquire operations performed during user
context first pin / last unpin.
When adding a request to a composite tracker, we try to reuse an existing
fence tracker, already allocated and registered with that composite. The
tracker we obtain may already track another fence, may be an idle barrier,
or an active barrier.
If the tracker we get occurs a non-idle barrier then we try to delete that
barrier from a list of barrier tasks it belongs to. However, while doing
that we don't respect return value from a function that performs the
barrier deletion. Should the deletion ever fail, we would end up reusing
the tracker still registered as a barrier task. Since the same structure
field is reused with both fence callback lists and barrier tasks list,
list corruptions would likely occur.
Barriers are now deleted from a barrier tasks list by temporarily removing
the list content, traversing that content with skip over the node to be
deleted, then populating the list back with the modified content. Should
that intentionally racy concurrent deletion attempts be not serialized,
one or more of those may fail because of the list being temporary empty.
Related code that ignores the results of barrier deletion was initially
introduced in v5.4 by commit b95b7a74051d ("drm/i915: Allow sharing the
idle-barrier from other kernel requests"). However, all users of the
barrier deletion routine were apparently serialized at that time, then the
issue didn't exhibit itself. Results of git bisect with help of a newly
developed igt@gem_barrier_race@remote-request IGT test indicate that list
corruptions might start to appear after commit fee3e9f60b02 ("drm/i915/gt:
Schedule request retirement when timeline idles"), introduced in v5.5.
Respect results of barrier deletion attempts -- mark the barrier as idle
only if successfully deleted from the list. Then, before proceeding with
setting our fence as the one currently tracked, make sure that the tracker
we've got is not a non-idle barrier. If that check fails then don't use
that tracker but go back and try to acquire a new, usable one.
v3: use unlikely() to document what outcome we expect (Andi),
- fix bad grammar in commit description.
v2: no code changes,
- blame commit fee3e9f60b02 ("drm/i915/gt: Schedule request retirement
when timeline idles"), v5.5, not commit b95b7a74051d ("drm/i915: Allow
sharing the idle-barrier from other kernel requests"), v5.4,
- reword commit description.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6333 Fixes: fee3e9f60b02 ("drm/i915/gt: Schedule request retirement when timeline idles") Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@vger.kernel.org # v5.5 Cc: Andi Shyti <andi.shyti@linux.intel.com> Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com> Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com> Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230302120820.48740-1-janusz.krzysztofik@linux.intel.com
(cherry picked from commit 7477a120c5be5e6da13eff3c949929544385450c) Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drm_gem_shmem_mmap() doesn't own reference in error code path, resulting
in the dma-buf shmem GEM object getting prematurely freed leading to a
later use-after-free.
After use_asid_allocator is enabled, the userspace application will
crash by stale TLB entries. Because only using cpumask_clear_cpu without
local_flush_tlb_all couldn't guarantee CPU's TLB entries were fresh.
Then set_mm_asid would cause the user space application to get a stale
value by stale TLB entry, but set_mm_noasid is okay.
Here is the symptom of the bug:
unhandled signal 11 code 0x1 (coredump)
0x0000003fd6d22524 <+4>: auipc s0,0x70
0x0000003fd6d22528 <+8>: ld s0,-148(s0) # 0x3fd6d92490
=> 0x0000003fd6d2252c <+12>: ld a5,0(s0)
(gdb) i r s0
s0 0x8082ed1cc3198b21 0x8082ed1cc3198b21
(gdb) x /2x 0x3fd6d92490
0x3fd6d92490: 0xd80ac8a8 0x0000003f
The core dump file shows that register s0 is wrong, but the value in
memory is correct. Because 'ld s0, -148(s0)' used a stale mapping entry
in TLB and got a wrong result from an incorrect physical address.
When the task ran on CPU0, which loaded/speculative-loaded the value of
address(0x3fd6d92490), then the first version of the mapping entry was
PTWed into CPU0's TLB.
When the task switched from CPU0 to CPU1 (No local_tlb_flush_all here by
asid), it happened to write a value on the address (0x3fd6d92490). It
caused do_page_fault -> wp_page_copy -> ptep_clear_flush ->
ptep_get_and_clear & flush_tlb_page.
The flush_tlb_page used mm_cpumask(mm) to determine which CPUs need TLB
flush, but CPU0 had cleared the CPU0's mm_cpumask in the previous
switch_mm. So we only flushed the CPU1 TLB and set the second version
mapping of the PTE. When the task switched from CPU1 to CPU0 again, CPU0
still used a stale TLB mapping entry which contained a wrong target
physical address. It raised a bug when the task happened to read that
value.
CPU0 CPU1
- switch 'task' in
- read addr (Fill stale mapping
entry into TLB)
- switch 'task' out (no tlb_flush)
- switch 'task' in (no tlb_flush)
- write addr cause pagefault
do_page_fault() (change to
new addr mapping)
wp_page_copy()
ptep_clear_flush()
ptep_get_and_clear()
& flush_tlb_page()
write new value into addr
- switch 'task' out (no tlb_flush)
- switch 'task' in (no tlb_flush)
- read addr again (Use stale
mapping entry in TLB)
get wrong value from old phyical
addr, BUG!
The solution is to keep all CPUs' footmarks of cpumask(mm) in switch_mm,
which could guarantee to invalidate all stale TLB entries during TLB
flush.
Samsung Galaxy Book2 Pro (13" 2022 NP930XED-KA1DE) with codec SSID
144d:c868 requires the same workaround for enabling the speaker amp
like other Samsung models with ALC298 code.
The effective values of the guest CR0 and CR4 registers may differ from
those included in the VMCS12. In particular, disabling EPT forces
CR4.PAE=1 and disabling unrestricted guest mode forces CR0.PG=CR0.PE=1.
Therefore, checks on these bits cannot be delegated to the processor
and must be performed by KVM.
Define AVIC_VCPU_ID_MASK based on AVIC_PHYSICAL_MAX_INDEX, i.e. the mask
that effectively controls the largest guest physical APIC ID supported by
x2AVIC, instead of hardcoding the number of bits to 8 (and the number of
VM bits to 24).
The AVIC GATag is programmed into the AMD IOMMU IRTE to provide a
reference back to KVM in case the IOMMU cannot inject an interrupt into a
non-running vCPU. In such a case, the IOMMU notifies software by creating
a GALog entry with the corresponded GATag, and KVM then uses the GATag to
find the correct VM+vCPU to kick. Dropping bit 8 from the GATag results
in kicking the wrong vCPU when targeting vCPUs with x2APIC ID > 255.
Define the "physical table max index mask" as bits 8:0, not 9:0. x2AVIC
currently supports a max of 512 entries, i.e. the max index is 511, and
the inputs to GENMASK_ULL() are inclusive. The bug is benign as bit 9 is
reserved and never set by KVM, i.e. KVM is just clearing bits that are
guaranteed to be zero.
Note, as of this writing, APM "Rev. 3.39-October 2022" incorrectly states
that bits 11:8 are reserved in Table B-1. VMCB Layout, Control Area. I.e.
that table wasn't updated when x2AVIC support was added.
Opportunistically fix the comment for the max AVIC ID to align with the
code, and clean up comment formatting too.
If the tracepoint is enabled, it could trigger RCU issues if called in
the wrong place. And this warning was only triggered if lockdep was
enabled. If the tracepoint was never enabled with lockdep, the bug would
not be caught. To handle this, the above sequence was done when lockdep
was enabled regardless if the tracepoint was enabled or not (although the
always enabled code really didn't do anything, it would still trigger a
warning).
But a lot has changed since that lockdep code was added. One is, that
sequence no longer triggers any warning. Another is, the tracepoint when
enabled doesn't even do that sequence anymore.
The main check we care about today is whether RCU is "watching" or not.
So if lockdep is enabled, always check if rcu_is_watching() which will
trigger a warning if it is not (tracepoints require RCU to be watching).
Note, that old sequence did add a bit of overhead when lockdep was enabled,
and with the latest kernel updates, would cause the system to slow down
enough to trigger kernel "stalled" warnings.
The function hist_field_name() cannot handle being passed a NULL field
parameter. It should never be NULL, but due to a previous bug, NULL was
passed to the function and the kernel crashed due to a NULL dereference.
Mark Rutland reported this to me on IRC.
The bug was fixed, but to prevent future bugs from crashing the kernel,
check the field and add a WARN_ON() if it is NULL.
Link: https://lkml.kernel.org/r/20230302020810.762384440@goodmis.org Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Reported-by: Mark Rutland <mark.rutland@arm.com> Fixes: 64363d35e3295 ("tracing: Add hist trigger 'sym' and 'sym-offset' modifiers") Tested-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>
Since the commit ec25677859f2 ("fs: don't allow splice read/write
without explicit ops") is applied to the kernel, splice() and
sendfile() calls on the trace file (/sys/kernel/debug/tracing
/trace) return EINVAL.
This patch restores these system calls by initializing splice_read
in file_operations of the trace file. This patch only enables such
functionalities for the read case.
Link: https://lore.kernel.org/linux-trace-kernel/20230314013707.28814-1-sfoon.kim@samsung.com Cc: stable@vger.kernel.org Fixes: ec25677859f2 ("fs: don't allow splice read/write without explicit ops") Signed-off-by: Sung-hun Kim <sfoon.kim@samsung.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Before my changes to how multichannel reconnects work, the
primary channel was always used to do a non-binding session
setup. With my changes, that is not the case anymore.
Missed this place where channel at index 0 was forcibly
updated with the signing key.
Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When BLOCK_LEGACY_AUTOLOAD is not enable, mdadm is not able to
activate new arrays unless "CREATE names=yes" appears in
mdadm.conf
As this is a regression we need to always enable BLOCK_LEGACY_AUTOLOAD
for when MD is selected - at least until mdadm is updated and the
updates widely available.
Cc: stable@vger.kernel.org # v5.18+ Fixes: bbe8c6dc275e ("block: deprecate autoloading based on dev_t") Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Song Liu <song@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to trigger a NULL-pointer
deference when either a NULL pointer or not fully initialised node is
returned from exynos_generic_icc_xlate().
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: 9d76140dcd1c ("interconnect: Add generic interconnect driver for Exynos SoCs") Cc: stable@vger.kernel.org # 5.11 Cc: Sylwester Nawrocki <s.nawrocki@samsung.com> Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-16-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: 4c2c9b495e3e ("interconnect: qcom: add msm8974 driver") Cc: stable@vger.kernel.org # 5.5 Reviewed-by: Brian Masney <bmasney@redhat.com> Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-12-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: b43753733b3f ("interconnect: qcom: Consolidate interconnect RPMh support") Cc: stable@vger.kernel.org # 5.7 Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-11-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: 7834eb78d031 ("interconnect: qcom: Consolidate interconnect RPM support") Fixes: d3a5d586798c ("interconnect: qcom: Add MSM8916 interconnect provider driver") Cc: stable@vger.kernel.org # 5.7 Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Reviewed-by: Jun Nie <jun.nie@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-9-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail:
of_icc_xlate_onecell: invalid index 0
cpu cpu0: error -EINVAL: error finding src node
cpu cpu0: dev_pm_opp_of_find_icc_paths: Unable to get path0: -22
qcom-cpufreq-hw: probe of 18591000.cpufreq failed with error -22
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: 0a20ef8382a0 ("interconnect: qcom: Add OSM L3 interconnect provider support") Cc: stable@vger.kernel.org # 5.7 Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-6-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
The current interconnect provider interface is inherently racy as
providers are expected to be added before being fully initialised.
Specifically, nodes are currently not added and the provider data is not
initialised until after registering the provider which can cause racing
DT lookups to fail.
Add a new provider API which will be used to fix up the interconnect
drivers.
The old API is reimplemented using the new interface and will be removed
once all drivers have been fixed.
The interconnect framework currently expects that providers are only
removed when there are no users and after all nodes have been removed.
There is currently nothing that guarantees this to be the case and the
framework does not do any reference counting, but refusing to remove the
provider is never correct as that would leave a dangling pointer to a
resource that is about to be released in the global provider list (e.g.
accessible through debugfs).
Replace the current sanity checks with WARN_ON() so that the provider is
always removed.
The code which handles the ipl report is searching for a free location
in memory where it could copy the component and certificate entries to.
It checks for intersection between the sections required for the kernel
and the component/certificate data area, but fails to check whether
the data structures linking these data areas together intersect.
This might cause the iplreport copy code to overwrite the iplreport
itself. Fix this by adding two addtional intersection checks.
Cc: <stable@vger.kernel.org> Fixes: 55590fd945be ("s390/ipl: read IPL report at early boot") Signed-off-by: Sven Schnelle <svens@linux.ibm.com> Reviewed-by: Vasily Gorbik <gor@linux.ibm.com> Signed-off-by: Vasily Gorbik <gor@linux.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The LRU mechanism may look up a resource in the process of being removed
from an object. The locking rules here are a bit unclear but it looks
currently like res->bo assignment is protected by the LRU lock, whereas
bo->resource is protected by the object lock, while *clearing* of
bo->resource is also protected by the LRU lock. This means that if
we check that bo->resource points to the LRU resource under the LRU
lock we should be safe.
So perform that check before deciding to swap out a bo. That avoids
dereferencing a NULL bo->resource in ttm_bo_swapout().
Fixes: 27a06b1672d8 ("drm/ttm: move the LRU into resource handling v4") Cc: Christian König <christian.koenig@amd.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Christian Koenig <christian.koenig@amd.com> Cc: Huang Rui <ray.huang@amd.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Felix Kuehling <Felix.Kuehling@amd.com> Cc: Philip Yang <Philip.Yang@amd.com> Cc: Qiang Yu <qiang.yu@amd.com> Cc: Matthew Auld <matthew.auld@intel.com> Cc: Nirmoy Das <nirmoy.das@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com> Cc: Anshuman Gupta <anshuman.gupta@intel.com> Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com> Cc: dri-devel@lists.freedesktop.org Cc: <stable@vger.kernel.org> # v5.19+ Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Reviewed-by: Christian König <christian.koenig@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230307144621.10748-2-thomas.hellstrom@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: a75551786b7b ("memory: tegra20: Support interconnect framework") Cc: stable@vger.kernel.org # 5.11 Cc: Dmitry Osipenko <digetx@gmail.com> Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-21-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: 851f5b26eb09 ("memory: tegra124: Support interconnect framework") Cc: stable@vger.kernel.org # 5.12 Cc: Dmitry Osipenko <digetx@gmail.com> Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-19-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
Fixes: a75551786b7b ("memory: tegra20: Support interconnect framework") Cc: stable@vger.kernel.org # 5.11 Cc: Dmitry Osipenko <digetx@gmail.com> Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20230306075651.2449-20-johan+linaro@kernel.org Signed-off-by: Georgi Djakov <djakov@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current interconnect provider registration interface is inherently
racy as nodes are not added until the after adding the provider. This
can specifically cause racing DT lookups to fail.
Switch to using the new API where the provider is not registered until
after it has been fully initialised.
REGMAP is a hidden (not user visible) symbol. Users cannot set it
directly thru "make *config", so drivers should select it instead of
depending on it if they need it.
Consistently using "select" or "depends on" can also help reduce
Kconfig circular dependency issues.
Therefore, change the use of "depends on REGMAP" to "select REGMAP".
The 8250 handle_irq callback is not just called from the interrupt
handler but also from a timer callback when polling (e.g. for ports
without an interrupt line). Consequently the callback must explicitly
disable interrupts to avoid a potential deadlock with another interrupt
in polled mode.
Fix up the two paths in the freescale callback that failed to re-enable
interrupts when polling.
As per HW manual for EMEV2 "R19UH0040EJ0400 Rev.4.00", the UART
IP found on EMMA mobile SoC is Register-compatible with the
general-purpose 16750 UART chip. Fix UART port type as 16750 and
enable 64-bytes fifo support.
According to LPUART RM, Transmission Complete Flag becomes 0 if queuing
a break character by writing 1 to CTRL[SBK], so here need to skip
waiting for transmission complete when UARTCTRL_SBK is asserted,
otherwise the kernel may stuck here.
And actually set_termios() adds transmission completion waiting to avoid
data loss or data breakage when changing the baud rate, but we don't
need to worry about this when queuing break characters.
[WHY]
When PTEBufferSizeInRequests is zero, UBSAN reports the following
warning because dml_log2 returns an unexpected negative value:
shift exponent 4294966273 is too large for 32-bit type 'int'
[HOW]
In the case PTEBufferSizeInRequests is zero, skip the dml_log2() and
assign the result directly.
Reviewed-by: Jun Lei <Jun.Lei@amd.com> Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com> Signed-off-by: Alex Hung <alex.hung@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The call trace occurs when the amdgpu is removed after
the mode1 reset. During mode1 reset, from suspend to resume,
there is no need to reinitialize the ta firmware buffer
which caused the bo pin_count increase redundantly.
GCC warns about the pattern sizeof(void*)/sizeof(void), as it looks like
the abuse of a pattern to calculate the array size. This pattern appears
in the unevaluated part of the ternary operator in _INTC_ARRAY if the
parameter is NULL.
The replacement uses an alternate approach to return 0 in case of NULL
which does not generate the pattern sizeof(void*)/sizeof(void), but still
emits the warning if _INTC_ARRAY is called with a nonarray parameter.
This patch is required for successful compilation with -Werror enabled.
The idea to use _Generic for type distinction is taken from Comment #7
in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108483 by Jakub Jelinek
Under CONFIG_DEBUG_ATOMIC_SLEEP=y and CONFIG_DEBUG_PREEMPT=y, we can see
the following messages on LoongArch, this is because using might_sleep()
in preemption disable context.
In order to avoid the above issue, we should break the call chains,
using timer_irq_installed variable as check condition to only call
get_timer_irq() once in constant_clockevent_init() is a simple and
proper way.
We are supposed to set fid->mode to reflect the flags
that were used to open the file. We were actually setting
it to the creation mode which is the default perms of the
file not the flags the file was opened with.
Signed-off-by: Eric Van Hensbergen <ericvh@kernel.org> Reviewed-by: Dominique Martinet <asmadeus@codewreck.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the kfd_wait_on_events() function, the kfd_event_waiter structure is
allocated by alloc_event_waiters(), but the event field of the waiter
structure is not initialized; When copy_from_user() fails in the
kfd_wait_on_events() function, it will enter exception handling to
release the previously allocated memory of the waiter structure;
Due to the event field of the waiters structure being accessed
in the free_waiters() function, this results in illegal memory access
and system crash, here is the crash log:
The problem is that the inode contains an xattr entry with ea_inum of 15
when cleaning up an orphan inode <15>. When evict inode <15>, the reference
counting of the corresponding EA inode is decreased. When EA inode <15> is
found by find_inode_fast() in __ext4_iget(), it is found that the EA inode
holds the I_FREEING flag and waits for the EA inode to complete deletion.
As a result, when inode <15> is being deleted, we wait for inode <15> to
complete the deletion, resulting in an infinite loop and triggering Hung
Task. To solve this problem, we only need to check whether the ino of EA
inode and parent is the same before getting EA inode.
When mounting a crafted ext4 image, s_journal_inum may change after journal
replay, which is obviously unreasonable because we have successfully loaded
and replayed the journal through the old s_journal_inum. And the new
s_journal_inum bypasses some of the checks in ext4_get_journal(), which
may trigger a null pointer dereference problem. So if s_journal_inum
changes after the journal replay, we ignore the change, and rewrite the
current journal_inum to the superblock.
In ext4_fill_super(), EXT4_ORPHAN_FS flag is cleared after
ext4_orphan_cleanup() is executed. Therefore, when __ext4_iget() is
called to get an inode whose i_nlink is 0 when the flag exists, no error
is returned. If the inode is a special inode, a null pointer dereference
may occur. If the value of i_nlink is 0 for any inodes (except boot loader
inodes) got by using the EXT4_IGET_SPECIAL flag, the current file system
is corrupted. Therefore, make the ext4_iget() function return an error if
it gets such an abnormal special inode.
The kernel disables all SSE and similar FP/SIMD instructions on
x86-based architectures (partly because we shouldn't be using floats in
the kernel, and partly to avoid the need for stack alignment, see:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53383 )
UML does not do the same thing, which isn't in itself a problem, but
does add to the list of differences between UML and "normal" x86 builds.
In addition, there was a crash bug with LLVM < 15 / rustc < 1.65 when
building with SSE, so disabling it fixes rust builds with earlier
compiler versions, see:
https://github.com/Rust-for-Linux/linux/pull/881
Signed-off-by: David Gow <davidgow@google.com> Reviewed-by: Sergio González Collado <sergio.collado@gmail.com> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
1. Write data to a file, say all 1s from offset 0 to 16.
2. Truncate the file to a smaller size, say 8 bytes.
3. Write new bytes (say 2s) from an offset past the original size of the
file, say at offset 20, for 4 bytes. This is supposed to create a "hole"
in the file, meaning that the bytes from offset 8 (where it was truncated
above) up to the new write at offset 20, should all be 0s (zeros).
4. Flush all caches using "echo 3 > /proc/sys/vm/drop_caches" (or unmount
and remount) the f/s.
5. Check the content of the file. It is wrong. The 1s that used to be
between bytes 9 and 16, before the truncation, have REAPPEARED (they should
be 0s).
We wrote a script and helper C program to reproduce the bug
(reproduce_jffs2_write_begin_issue.sh, write_file.c, and Makefile). We can
make them available to anyone.
The above example is shown when writing a small file within the same first
page. But the bug happens for larger files, as long as steps 1, 2, and 3
above all happen within the same page.
The problem was traced to the jffs2_write_begin code, where it goes into an
'if' statement intended to handle writes past the current EOF (i.e., writes
that may create a hole). The code computes a 'pageofs' that is the floor
of the write position (pos), aligned to the page size boundary. In other
words, 'pageofs' will never be larger than 'pos'. The code then sets the
internal jffs2_raw_inode->isize to the size of max(current inode size,
pageofs) but that is wrong: the new file size should be the 'pos', which is
larger than both the current inode size and pageofs.
Similarly, the code incorrectly sets the internal jffs2_raw_inode->dsize to
the difference between the pageofs minus current inode size; instead it
should be the current pos minus the current inode size. Finally,
inode->i_size was also set incorrectly.
The patch below fixes this bug. The bug was discovered using a new tool
for finding f/s bugs using model checking, called MCFS (Model Checking File
Systems).
Signed-off-by: Yifei Liu <yifeliu@cs.stonybrook.edu> Signed-off-by: Erez Zadok <ezk@cs.stonybrook.edu> Signed-off-by: Manish Adkar <madkar@cs.stonybrook.edu> Signed-off-by: Richard Weinberger <richard@nod.at> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some projects might not support CONFIG_DEBUG_FS but still needs svs to be
alive. Therefore, enclose debug cmd codes with CONFIG_DEBUG_FS to make sure
svs can be alive when CONFIG_DEBUG_FS not supported.
This commit fixes a race between completion of stop command and start of a
new command.
Previously the command ready interrupt was enabled before stop command
was written to the command register. This caused the command ready
interrupt to fire immediately since the CMDRDY flag is asserted constantly
while there is no command in progress.
Consequently the command state machine will immediately advance to the
next state when the tasklet function is executed again, no matter
actual completion state of the stop command.
Thus a new command can then be dispatched immediately, interrupting and
corrupting the stop command on the CMD line.
Fix that by dropping the command ready interrupt enable before calling
atmci_send_stop_cmd. atmci_send_stop_cmd does already enable the
command ready interrupt, no further writes to ATMCI_IER are necessary.
The __find_restype() function loops over the m5mols_default_ffmt[]
array, and the termination condition ends up being wrong: instead of
stopping when the iterator becomes the size of the array it traverses,
it stops after it has already overshot the array.
Now, in practice this doesn't likely matter, because the code will
always find the entry it looks for, and will thus return early and never
hit that last extra iteration.
But it turns out that clang will unroll the loop fully, because it has
only two iterations (well, three due to the off-by-one bug), and then
clang will end up just giving up in the middle of the loop unrolling
when it notices that the code walks past the end of the array.
And that made 'objtool' very unhappy indeed, because the generated code
just falls off the edge of the universe, and ends up falling through to
the next function, causing this warning:
drivers/media/i2c/m5mols/m5mols.o: warning: objtool: m5mols_set_fmt() falls through to next function m5mols_get_frame_desc()
Fix the loop ending condition.
Reported-by: Jens Axboe <axboe@kernel.dk> Analyzed-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Analyzed-by: Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/linux-block/CAHk-=wgTSdKYbmB1JYM5vmHMcD9J9UZr0mn7BOYM_LudrP+Xvw@mail.gmail.com/ Fixes: f3e08500ecb8 ("[media] Add support for M-5MOLS 8 Mega Pixel camera ISP") Cc: HeungJun, Kim <riverful.kim@samsung.com> Cc: Sylwester Nawrocki <s.nawrocki@samsung.com> Cc: Kyungmin Park <kyungmin.park@samsung.com> Cc: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ltc2992 drivers uses a mutex and I2C bus access in its GPIO chip `set`
and `get` implementation. This means these functions can sleep and the GPIO
chip should set the `can_sleep` property to true.
This will ensure that a warning is printed when trying to set or get the
GPIO value from a context that potentially can't sleep.
The adm1266 driver uses I2C bus access in its GPIO chip `set` and `get`
implementation. This means these functions can sleep and the GPIO chip
should set the `can_sleep` property to true.
This will ensure that a warning is printed when trying to set or get the
GPIO value from a context that potentially can't sleep.
Prior to commit 8d17b7d0d7fa ("kconfig: change sym_change_count to a
boolean flag"), the conf_updated flag was set to the new value *before*
calling the callback. xconfig's save action depends on this behaviour,
because xconfig calls conf_get_changed() directly from the callback and
now sees the old value, thus never enabling the save button or the
shortcut.
Restore the previous behaviour.
Fixes: 8d17b7d0d7fa ("kconfig: change sym_change_count to a boolean flag") Signed-off-by: Jurica Vukadin <jura@vukad.in> Acked-by: Randy Dunlap <rdunlap@infradead.org> Tested-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The driver will match mostly by DT table (even thought there is regular
ID table) so there is little benefit in of_match_ptr (this also allows
ACPI matching via PRP0001, even though it might not be relevant here).
This also fixes !CONFIG_OF error:
drivers/hwmon/tmp513.c:610:34: error: ‘tmp51x_of_match’ defined but not used [-Werror=unused-const-variable=]
When probing the ucd90320 access to some of the registers randomly fails.
Sometimes it NACKs a transfer, sometimes it returns just random data and
the PEC check fails.
Experimentation shows that this seems to be triggered by a register access
directly back to back with a previous register write. Experimentation also
shows that inserting a small delay after register writes makes the issue go
away.
Use a similar solution to what the max15301 driver does to solve the same
problem. Create a custom set of bus read and write functions that make sure
that the delay is added.
Fixes: 8c2d01088b85 ("hwmon: (pmbus/ucd9000) Add support for UCD90320 Power Sequencer") Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Link: https://lore.kernel.org/r/20230312160312.2227405-1-lars@metafoo.de Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The wrong bits are masked in the hysteresis register; indices 0 and 2
should zero bits [7:4] and preserve bits [3:0], and index 1 should zero
bits [3:0] and preserve bits [7:4].
Fixes: 9710f8bd109c ("hwmon: Add a driver for the ADT7475 hardware monitoring chip") Signed-off-by: Tony O'Brien <tony.obrien@alliedtelesis.co.nz> Link: https://lore.kernel.org/r/20230222005228.158661-3-tony.obrien@alliedtelesis.co.nz Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Throughout the ADT7475 driver, attributes relating to the temperature
sensors are displayed in the order Remote 1, Local, Remote 2. Make
temp_st_show() conform to this expectation so that values set by
temp_st_store() can be displayed using the correct attribute.
syzbot reported a warning[1] where the bond device itself is a slave and
we try to enslave a non-ethernet device as the first slave which fails
but then in the error path when ether_setup() restores the bond device
it also clears all flags. In my previous fix[2] I restored the
IFF_MASTER flag, but I didn't consider the case that the bond device
itself might also be a slave with IFF_SLAVE set, so we need to restore
that flag as well. Use the bond_ether_setup helper which does the right
thing and restores the bond's flags properly.
Steps to reproduce using a nlmon dev:
$ ip l add nlmon0 type nlmon
$ ip l add bond1 type bond
$ ip l add bond2 type bond
$ ip l set bond1 master bond2
$ ip l set dev nlmon0 master bond1
$ ip -d l sh dev bond1
22: bond1: <BROADCAST,MULTICAST,MASTER> mtu 1500 qdisc noqueue master bond2 state DOWN mode DEFAULT group default qlen 1000
(now bond1's IFF_SLAVE flag is gone and we'll hit a warning[3] if we
try to delete it)
Fixes: 1dbd6db88310 ("bonding: correctly handle bonding type change on enslave failure") Reported-by: syzbot+9dfc3f3348729cc82277@syzkaller.appspotmail.com Link: https://syzkaller.appspot.com/bug?id=391c7b1f6522182899efba27d891f1743e8eb3ef Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org> Reviewed-by: Michal Kubiak <michal.kubiak@intel.com> Acked-by: Jonathan Toppins <jtoppins@redhat.com> Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Add bond_ether_setup helper which is used to fix ether_setup() calls in the
bonding driver. It takes care of both IFF_MASTER and IFF_SLAVE flags, the
former is always restored and the latter only if it was set.
If the bond enslaves non-ARPHRD_ETHER device (changes its type), then
releases it and enslaves ARPHRD_ETHER device (changes back) then we
use ether_setup() to restore the bond device type but it also resets its
flags and removes IFF_MASTER and IFF_SLAVE[1]. Use the bond_ether_setup
helper to restore both after such transition.
[1] reproduce (nlmon is non-ARPHRD_ETHER):
$ ip l add nlmon0 type nlmon
$ ip l add bond2 type bond mode active-backup
$ ip l set nlmon0 master bond2
$ ip l set nlmon0 nomaster
$ ip l add bond1 type bond
(we use bond1 as ARPHRD_ETHER device to restore bond2's mode)
$ ip l set bond1 master bond2
$ ip l sh dev bond2
37: bond2: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether be:d7:c5:40:5b:cc brd ff:ff:ff:ff:ff:ff promiscuity 0 minmtu 68 maxmtu 1500
(notice bond2's IFF_MASTER is missing)
Fixes: 3e239f541fa4 ("bonding: clean muticast addresses when device changes type") Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The blamed commit has replaced a ksz_write8() call to address
REG_PORT_5_CTRL_6 (0x56) with a ksz_set_xmii() -> ksz_pwrite8() call to
regs[P_XMII_CTRL_1], which is also defined as 0x56 for ksz8795_regs[].
The trouble is that, when compared to ksz_write8(), ksz_pwrite8() also
adjusts the register offset with the port base address. So in reality,
ksz_pwrite8(offset=0x56) accesses register 0x56 + 0x50 = 0xa6, which in
this switch appears to be unmapped, and the RGMII delay configuration on
the CPU port does nothing.
So if the switch wasn't fine with the RGMII delay configuration done
through pin strapping and relied on Linux to apply a different one in
order to pass traffic, this is now broken.
Using the offset translation logic imposed by ksz_pwrite8(), the correct
value for regs[P_XMII_CTRL_1] should have been 0x6 on ksz8795_regs[], in
order to really end up accessing register 0x56.
Static code analysis shows that, despite there being multiple other
accesses to regs[P_XMII_CTRL_1] in this driver, the only code path that
is applicable to ksz8795_regs[] and ksz8_dev_ops is ksz_set_xmii().
Therefore, the problem is isolated to RGMII delays.
In its current form, ksz8795_regs[] contains the same value for
P_XMII_CTRL_0 and for P_XMII_CTRL_1, and this raises valid suspicions
that writes made by the driver to regs[P_XMII_CTRL_0] might overwrite
writes made to regs[P_XMII_CTRL_1] or vice versa.
Again, static analysis shows that the only accesses to P_XMII_CTRL_0
from the driver are made from code paths which are not reachable with
ksz8_dev_ops. So the accesses made by ksz_set_xmii() are safe for this
switch family.
[ vladimiroltean: rewrote commit message ]
Fixes: 30e80099244f ("net: dsa: microchip: ksz8795: use common xmii function") Signed-off-by: Marek Vasut <marex@denx.de> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Acked-by: Arun Ramadoss <arun.ramadoss@microchip.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Link: https://lore.kernel.org/r/20230315231916.2998480-1-vladimir.oltean@nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The `devlink -j port show` command output may not contain the "flavour"
key, an example from Ubuntu 22.10 s390x LPAR(5.19.0-37-generic), with
mlx4 driver and iproute2-5.15.0:
{"port":{"pci/0001:00:00.0/1":{"type":"eth","netdev":"ens301"},
"pci/0001:00:00.0/2":{"type":"eth","netdev":"ens301d1"},
"pci/0002:00:00.0/1":{"type":"eth","netdev":"ens317"},
"pci/0002:00:00.0/2":{"type":"eth","netdev":"ens317d1"}}}
This will cause a KeyError exception.
Create a validate_devlink_output() to check for this "flavour" from
devlink command output to avoid this KeyError exception. Also let
it handle the check for `devlink -j dev show` output in main().
Apart from this, if the test was not started because the max lanes of
the designated device is 0. The script will still return 0 and thus
causing a false-negative test result.
Use a found_max_lanes flag to determine if these tests were skipped
due to this reason and return KSFT_SKIP to make it more clear.
iucv_irq_data needs to be 4 bytes larger.
These bytes are not used by the iucv module, but written by
the z/VM hypervisor in case a CPU is deconfigured.
Reported as:
BUG dma-kmalloc-64 (Not tainted): kmalloc Redzone overwritten
-----------------------------------------------------------------------------
0x0000000000400564-0x0000000000400567 @offset=1380. First byte 0x80 instead of 0xcc
Allocated in iucv_cpu_prepare+0x44/0xd0 age=167839 cpu=2 pid=1
__kmem_cache_alloc_node+0x166/0x450
kmalloc_node_trace+0x3a/0x70
iucv_cpu_prepare+0x44/0xd0
cpuhp_invoke_callback+0x156/0x2f0
cpuhp_issue_call+0xf0/0x298
__cpuhp_setup_state_cpuslocked+0x136/0x338
__cpuhp_setup_state+0xf4/0x288
iucv_init+0xf4/0x280
do_one_initcall+0x78/0x390
do_initcalls+0x11a/0x140
kernel_init_freeable+0x25e/0x2a0
kernel_init+0x2e/0x170
__ret_from_fork+0x3c/0x58
ret_from_fork+0xa/0x40
Freed in iucv_init+0x92/0x280 age=167839 cpu=2 pid=1
__kmem_cache_free+0x308/0x358
iucv_init+0x92/0x280
do_one_initcall+0x78/0x390
do_initcalls+0x11a/0x140
kernel_init_freeable+0x25e/0x2a0
kernel_init+0x2e/0x170
__ret_from_fork+0x3c/0x58
ret_from_fork+0xa/0x40
Slab 0x0000037200010000 objects=32 used=30 fp=0x0000000000400640 flags=0x1ffff00000010200(slab|head|node=0|zone=0|
Object 0x0000000000400540 @offset=1344 fp=0x0000000000000000
Redzone 0000000000400500: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Redzone 0000000000400510: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Redzone 0000000000400520: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Redzone 0000000000400530: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Object 0000000000400540: 00 01 00 03 00 00 00 00 00 00 00 00 00 00 00 00 ................
Object 0000000000400550: f3 86 81 f2 f4 82 f8 82 f0 f0 f0 f0 f0 f0 f0 f2 ................
Object 0000000000400560: 00 00 00 00 80 00 00 00 cc cc cc cc cc cc cc cc ................
Object 0000000000400570: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc ................
Redzone 0000000000400580: cc cc cc cc cc cc cc cc ........
Padding 00000000004005d4: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
Padding 00000000004005e4: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZZZZZ
Padding 00000000004005f4: 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZZZZZ
CPU: 6 PID: 121030 Comm: 116-pai-crypto. Not tainted 6.3.0-20230221.rc0.git4.99b8246b2d71.300.fc37.s390x+debug #1
Hardware name: IBM 3931 A01 704 (z/VM 7.3.0)
Call Trace:
[<000000032aa034ec>] dump_stack_lvl+0xac/0x100
[<0000000329f5a6cc>] check_bytes_and_report+0x104/0x140
[<0000000329f5aa78>] check_object+0x370/0x3c0
[<0000000329f5ede6>] free_debug_processing+0x15e/0x348
[<0000000329f5f06a>] free_to_partial_list+0x9a/0x2f0
[<0000000329f5f4a4>] __slab_free+0x1e4/0x3a8
[<0000000329f61768>] __kmem_cache_free+0x308/0x358
[<000000032a91465c>] iucv_cpu_dead+0x6c/0x88
[<0000000329c2fc66>] cpuhp_invoke_callback+0x156/0x2f0
[<000000032aa062da>] _cpu_down.constprop.0+0x22a/0x5e0
[<0000000329c3243e>] cpu_device_down+0x4e/0x78
[<000000032a61dee0>] device_offline+0xc8/0x118
[<000000032a61e048>] online_store+0x60/0xe0
[<000000032a08b6b0>] kernfs_fop_write_iter+0x150/0x1e8
[<0000000329fab65c>] vfs_write+0x174/0x360
[<0000000329fab9fc>] ksys_write+0x74/0x100
[<000000032aa03a5a>] __do_syscall+0x1da/0x208
[<000000032aa177b2>] system_call+0x82/0xb0
INFO: lockdep is turned off.
FIX dma-kmalloc-64: Restoring kmalloc Redzone 0x0000000000400564-0x0000000000400567=0xcc
FIX dma-kmalloc-64: Object at 0x0000000000400540 not freed
Fixes: 711b3886e693 ("[S390]: Rewrite of the IUCV base code, part 2") Signed-off-by: Alexandra Winter <wintera@linux.ibm.com> Link: https://lore.kernel.org/r/20230315131435.4113889-1-wintera@linux.ibm.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The aq_xdp_run_prog() function falls back to the XDP_ABORTED action
handler (using a goto) if the operations for any of the other actions fail.
The XDP_ABORTED handler in turn calls the bpf_warn_invalid_xdp_action()
tracepoint. However, the function also jumps into the XDP_PASS helper if no
XDP program is loaded on the device, which means the XDP_ABORTED handler
can be run with a NULL program pointer. This results in a NULL pointer
deref because the tracepoint dereferences the 'prog' pointer passed to it.
This situation can happen in multiple ways:
- If a packet arrives between the removal of the program from the interface
and the static_branch_dec() in aq_xdp_setup()
- If there are multiple devices using the same driver in the system and
one of them has an XDP program loaded and the other does not.
Fix this by refactoring the aq_xdp_run_prog() function to remove the 'goto
pass' handling if there is no XDP program loaded. Instead, factor out the
skb building in a separate small helper function.
Packet length check needs to be located after size and align_count
calculation to prevent kernel panic in skb_pull() in case
rx_cmd_a & RX_CMD_A_RED evaluates to true.
Commit 684c478bb28d ("ipv4: Fix incorrect route flushing when source
address is deleted") started to take the table ID field in the FIB info
structure into account when determining if two structures are identical
or not. This field is initialized using the 'fc_table' field in the
route configuration structure, which is not set when adding a route via
IOCTL.
The above can result in user space being able to install two identical
routes that only differ in the table ID field of their associated FIB
info.
Fix by initializing the table ID field in the route configuration
structure in the IOCTL path.
Before the fix:
# ip route add default via 192.0.2.2
# route add default gw 192.0.2.2
# ip -4 r show default
# default via 192.0.2.2 dev dummy10
# default via 192.0.2.2 dev dummy10
After the fix:
# ip route add default via 192.0.2.2
# route add default gw 192.0.2.2
SIOCADDRT: File exists
# ip -4 r show default
default via 192.0.2.2 dev dummy10
Audited the code paths to ensure there are no other paths that do not
properly initialize the route configuration structure when installing a
route.
Fixes: b9ef42ef6125 ("net: Don't delete routes in different VRFs") Fixes: 684c478bb28d ("ipv4: Fix incorrect route flushing when source address is deleted") Reported-by: gaoxingwang <gaoxingwang1@huawei.com> Link: https://lore.kernel.org/netdev/20230314144159.2354729-1-gaoxingwang1@huawei.com/ Tested-by: gaoxingwang <gaoxingwang1@huawei.com> Signed-off-by: Ido Schimmel <idosch@nvidia.com> Reviewed-by: David Ahern <dsahern@kernel.org> Link: https://lore.kernel.org/r/20230315124009.4015212-1-idosch@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
SH_ETH doesn't need mdiobus suspend/resume, that's why it sets
'mac_managed_pm'. However, setting it needs to be moved from init to
probe, so mdiobus PM functions will really never be called (e.g. when
the interface is not up yet during suspend/resume).
Fixes: cf2e91979a0a ("net: sh_eth: Fix PHY state warning splat during system resume") Suggested-by: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Michal Kubiak <michal.kubiak@intel.com> Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
RAVB doesn't need mdiobus suspend/resume, that's why it sets
'mac_managed_pm'. However, setting it needs to be moved from init to
probe, so mdiobus PM functions will really never be called (e.g. when
the interface is not up yet during suspend/resume).
Fixes: 45902c1775be ("net: ravb: Fix PHY state warning splat during system resume") Suggested-by: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Michal Kubiak <michal.kubiak@intel.com> Reviewed-by: Sergey Shtylyov <s.shtylyov@omp.ru> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There are 3 classes of switch families that the driver is aware of, as
far as mv88e6xxx_change_mtu() is concerned:
- MTU configuration is available per port. Here, the
chip->info->ops->port_set_jumbo_size() method will be present.
- MTU configuration is global to the switch. Here, the
chip->info->ops->set_max_frame_size() method will be present.
- We don't know how to change the MTU. Here, none of the above methods
will be present.
Switch families MV88E6165, MV88E6191, MV88E6220, MV88E6250 and MV88E6290
fall in category 3.
The blamed commit has adjusted the MTU for all 3 categories by EDSA_HLEN
(8 bytes), resulting in a new maximum MTU of 1492 being reported by the
driver for these switches.
I don't have the hardware to test, but I do have a MV88E6390 switch on
which I can simulate this by commenting out its .port_set_jumbo_size
definition from mv88e6390_ops. The result is this set of messages at
probe time:
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 1
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 2
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 3
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 4
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 5
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 6
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 7
mv88e6085 d0032004.mdio-mii:10: nonfatal error -34 setting MTU to 1500 on port 8
It is highly implausible that there exist Ethernet switches which don't
support the standard MTU of 1500 octets, and this is what the DSA
framework says as well - the error comes from dsa_slave_create() ->
dsa_slave_change_mtu(slave_dev, ETH_DATA_LEN).
But the error messages are alarming, and it would be good to suppress
them.
As a consequence of this unlikeliness, we reimplement mv88e6xxx_get_max_mtu()
and mv88e6xxx_change_mtu() on switches from the 3rd category as follows:
the maximum supported MTU is 1500, and any request to set the MTU to a
value larger than that fails in dev_validate_mtu().
Fixes: ea2f9182a4cd ("dsa: mv88e6xxx: Include tagger overhead when setting MTU for DSA and CPU ports") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, when dsa_slave_change_mtu() is called on a user port where
dev->max_mtu is 1500 (as returned by ds->ops->port_max_mtu()), the code
will stumble upon this check:
if (new_master_mtu > mtu_limit)
return -ERANGE;
because new_master_mtu is adjusted for the tagger overhead but mtu_limit
is not.
But it would be good if the logic went through, for example if the DSA
master really depends on an MTU adjustment to accept DSA-tagged frames.
To make the code pass through the check, we need to adjust mtu_limit for
the overhead as well, if the minimum restriction was caused by the DSA
user port's MTU (dev->max_mtu). A DSA user port MTU and a DSA master MTU
are always offset by the protocol overhead.
Currently no drivers return 1500 .port_max_mtu(), but this is only
temporary and a bug in itself - mv88e6xxx should have done that, but
since commit ea2f9182a4cd ("dsa: mv88e6xxx: Include tagger overhead when
setting MTU for DSA and CPU ports") it no longer does. This is a
preparation for fixing that.
Fixes: b202c04f8d0d ("net: dsa: configure the MTU for switch ports") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
ice_qp_dis() intends to stop a given queue pair that is a target of xsk
pool attach/detach. One of the steps is to disable interrupts on these
queues. It currently is broken in a way that txq irq is turned off
*after* HW flush which in turn takes no effect.
Below splat can be triggered by following steps:
- start xdpsock WITHOUT loading xdp prog
- run xdp_rxq_info with XDP_TX action on this interface
- start traffic
- terminate xdpsock
In fact, irqs were not disabled and napi managed to be scheduled and run
while xsk_pool pointer was still valid, but SW ring of xdp_buff pointers
was already freed.
To fix this, call ice_qvec_dis_irq() after ice_vsi_stop_tx_ring(). Also
while at it, remove redundant ice_clean_rx_ring() call - this is handled
in ice_qp_clean_rings().
Fixes: a40d09f79c37 ("ice: Add support for AF_XDP") Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Reviewed-by: Larysa Zaremba <larysa.zaremba@intel.com> Tested-by: Chandan Kumar Rout <chandanx.rout@intel.com> (A Contingent Worker at Intel) Acked-by: John Fastabend <john.fastabend@gmail.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 519e3d5b6256 ("veth: Rework veth_xdp_rcv_skb in order
to accept non-linear skb") introduced a bug where it tried to
use pskb_expand_head() if the headroom was less than
XDP_PACKET_HEADROOM. This however uses kmalloc to expand the head,
which will later allow consume_skb() to free the skb while is it still
in use by AF_XDP.
Previously if the headroom was less than XDP_PACKET_HEADROOM we
continued on to allocate a new skb from pages so this restores that
behavior.
BUG: KASAN: use-after-free in __xsk_rcv+0x18d/0x2c0
Read of size 78 at addr ffff888976250154 by task napi/iconduit-g/148640
The buggy address belongs to the object at ffff888976250000
which belongs to the cache kmalloc-2k of size 2048
The buggy address is located 340 bytes inside of
2048-byte region [ffff888976250000, ffff888976250800)
Currently, BREAK_FW_WAIT flag is set after syncing with fw_reset.
However, fw_reset can call mlx5_load_one() which is waiting for fw
init bit and BREAK_FW_WAIT flag is intended to stop. e.g.: the driver
might wait on a loop it should exit.
Fix it by setting the flag before syncing with fw_reset.
During module is unloaded while a peer tc flow is still offloaded,
first the peer uplink rep profile is changed to a nic profile, and so
neigh encap lock is destroyed. Next during unload, the VF reps netdevs
are unregistered which causes the original non-peer tc flow to be deleted,
which deletes the peer flow. The peer flow deletion detaches the encap
entry and try to take the already destroyed encap lock, causing the
below trace.
Fix this by clearing peer flows during tc eswitch cleanup
(mlx5e_tc_esw_cleanup()).
Rules with mirror actions are split to two FTEs when the actions after the mirror
action contains pedit, vlan push/pop or ct. Forward to ovs internal port adds
implicit header rewrite (pedit) but missing trigger to do split.
Fix by setting split_count when forwarding to ovs internal port which
will trigger split in mirror rules.
Fixes: 2fa4ad35f7b2 ("net/mlx5e: Offload tc rules that redirect to ovs internal port") Signed-off-by: Maor Dickman <maord@nvidia.com> Reviewed-by: Roi Dayan <roid@nvidia.com> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In few cases, rules with mirror use case are split to two FTEs, one which
do the mirror action and forward to second FTE which do the rest of the rule
actions and the second redirect action.
In case of mirror rules which do split and forward to ovs internal port or
VF stack devices, source port rewrite should be used in the second FTE but
it is wrongly also set in the first FTE which break the offload.
Fix this issue by removing the wrong check if source port rewrite is needed to
be used on the first FTE of the split and instead return EOPNOTSUPP which will
block offload of rules which mirror to ovs internal port or VF stack devices
which isn't supported.
projects / kernel.git / log