Testing shown that when a wq mode is setup to be dedicated and then torn
down and reconfigured to shared, the wq configured end up being dedicated
anyays. The root cause is when idxd_device_wqs_clear_state() gets called
during idxd_driver removal, idxd_wq_disable_cleanup() does not get called
vs when the wq driver is removed first. The check of wq state being
"enabled" causes the cleanup to be bypassed. However, idxd_driver->remove()
releases all wq drivers. So the wqs goes to "disabled" state and will never
be "enabled". By that point, the driver has no idea if the wq was
previously configured or clean. So force call idxd_wq_disable_cleanup() on
all wqs always to make sure everything gets cleaned up.
Reported-by: Tony Zhu <tony.zhu@intel.com> Tested-by: Tony Zhu <tony.zhu@intel.com> Fixes: 0dcfe41e9a4c ("dmanegine: idxd: cleanup all device related bits after disabling device") Signed-off-by: Dave Jiang <dave.jiang@intel.com> Co-developed-by: Fenghua Yu <fenghua.yu@intel.com> Signed-off-by: Fenghua Yu <fenghua.yu@intel.com> Link: https://lore.kernel.org/r/20220628230056.2527816-1-fenghua.yu@intel.com Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit dbad41e7bb5f ("dmaengine: qcom: bam_dma: check if the runtime pm enabled")
caused unbalanced pm_runtime_get/put() calls when the bam is
controlled remotely. This commit reverts it and just enables pm_runtime
in all cases, the clk_* functions already just nop when the clock is NULL.
Also clean up a bit by removing unnecessary bamclk null checks.
Suggested-by: Stephan Gerhold <stephan@gerhold.net> Fixes: dbad41e7bb5f ("dmaengine: qcom: bam_dma: check if the runtime pm enabled") Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org> Link: https://lore.kernel.org/r/20220629140559.118537-1-caleb.connolly@linaro.org Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It seems that it is valid to have less than the requested number of
descriptors. But what is not valid and leads to subsequent errors is to
have zero descriptors. In that case, abort the probing.
Fixes: e1f7c9eee707 ("dmaengine: at_xdmac: creation of the atmel eXtended DMA Controller driver") Signed-off-by: Michael Walle <michael@walle.cc> Link: https://lore.kernel.org/r/20220526135111.1470926-1-michael@walle.cc Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The DEFINE_SPINLOCK() macro shouldn't be used for dynamically allocated
spinlocks. The lockdep warns about this and disables locking validator.
Fix the warning by making lock static.
INFO: trying to register non-static key.
The code is fine but needs lockdep annotation, or maybe
you didn't initialize this object before use?
turning off the locking correctness validator.
Hardware name: Radxa ROCK Pi 4C (DT)
Call trace:
dump_backtrace.part.0+0xcc/0xe0
show_stack+0x18/0x6c
dump_stack_lvl+0x8c/0xb8
dump_stack+0x18/0x34
register_lock_class+0x4a8/0x4cc
__lock_acquire+0x78/0x20cc
lock_acquire.part.0+0xe0/0x230
lock_acquire+0x68/0x84
_raw_spin_lock_irqsave+0x84/0xc4
add_desc+0x44/0xc0
pl330_get_desc+0x15c/0x1d0
pl330_prep_dma_cyclic+0x100/0x270
snd_dmaengine_pcm_trigger+0xec/0x1c0
dmaengine_pcm_trigger+0x18/0x24
...
This is another old BUG_ON() that just shouldn't exist (see also commit a382f8fee42c: "signal handling: don't use BUG_ON() for debugging").
In fact, as Matthew Wilcox points out, this condition shouldn't really
even result in a warning, since a negative id allocation result is just
a normal allocation failure:
"I wonder if we should even warn here -- sure, the caller is trying to
free something that wasn't allocated, but we don't warn for
kfree(NULL)"
and goes on to point out how that current error check is only causing
people to unnecessarily do their own index range checking before freeing
it.
This was noted by Itay Iellin, because the bluetooth HCI socket cookie
code does *not* do that range checking, and ends up just freeing the
error case too, triggering the BUG_ON().
The HCI code requires CAP_NET_RAW, and seems to just result in an ugly
splat, but there really is no reason to BUG_ON() here, and we have
generally striven for allocation models where it's always ok to just do
free(alloc());
even if the allocation were to fail for some random reason (usually
obviously that "random" reason being some resource limit).
Fixes: 88eca0207cf1 ("ida: simplified functions for id allocation") Reported-by: Itay Iellin <ieitayie@gmail.com> Suggested-by: Matthew Wilcox <willy@infradead.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The conditional block for variants with a second clock should have set
minItems, not maxItems, which was already 2. Since clock-names requires
two items, this typo should not have caused any problems.
Fixes: edd14218bd66 ("dt-bindings: dmaengine: Convert Allwinner A31 and A64 DMA to a schema") Signed-off-by: Samuel Holland <samuel@sholland.org> Reviewed-by: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20220702031903.21703-1-samuel@sholland.org Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the DMA is configured for more than 8 channels the bits controlling
suspend moves to another register. However when adding support for this
the new register would be completely overwritten in one case and
overwritten with values from the old register in another case.
Found by comparing the parallel implementation of more than 8 channel
support for the StarFive JH7100 SoC by Samin.
Set return value in rsp_buf alloc error path before going to
error handling.
drivers/misc/cardreader/rtsx_usb.c:639:6: warning: variable 'ret' is used uninitialized whenever 'if' condition is true [-Wsometimes-uninitialized]
if (!ucr->rsp_buf)
^~~~~~~~~~~~~
drivers/misc/cardreader/rtsx_usb.c:678:9: note: uninitialized use occurs here
return ret;
^~~
drivers/misc/cardreader/rtsx_usb.c:639:2: note: remove the 'if' if its condition is always false
if (!ucr->rsp_buf)
^~~~~~~~~~~~~~~~~~
drivers/misc/cardreader/rtsx_usb.c:622:9: note: initialize the variable 'ret' to silence this warning
int ret;
^
= 0
Fixes: 3776c7855985 ("misc: rtsx_usb: use separate command and response buffers") Reported-by: kernel test robot <lkp@intel.com> Cc: stable <stable@kernel.org> Signed-off-by: Shuah Khan <skhan@linuxfoundation.org> Link: https://lore.kernel.org/r/20220701165352.15687-1-skhan@linuxfoundation.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
rtsx_usb uses same buffer for command and response. There could
be a potential conflict using the same buffer for both especially
if retries and timeouts are involved.
Use separate command and response buffers to avoid conflicts.
rtsx_usb driver allocates coherent dma buffer for urb transfers.
This buffer is passed to usb_bulk_msg() and usb core tries to
map already mapped buffer running into a dma mapping error.
xhci_hcd 0000:01:00.0: rejecting DMA map of vmalloc memory
WARNING: CPU: 1 PID: 279 at include/linux/dma-mapping.h:326 usb_ hcd_map_urb_for_dma+0x7d6/0x820
An interrupt for a channel might be pending even after struct
dma_device::device_terminate_all has been called. In that case the
recently introduced warning message "restart cyclic channel..." triggers
and the channel will be restarted. This is not desired as the channel
has just been stopped. Only restart the channel when we still have a
descriptor set for it (which will be set to NULL in
sdma_terminate_all()).
The revision of the imx-sdma IP that is in the i.MX8M series is the
same is that as that in the i.MX7 series but the imx7d MODULE_FIRMWARE
directive is wrapped in a condiditional which means it's not defined
when built for aarch64 SOC_IMX8M platforms and hence you get the
following errors when the driver loads on imx8m devices:
imx-sdma 302c0000.dma-controller: Direct firmware load for imx/sdma/sdma-imx7d.bin failed with error -2
imx-sdma 302c0000.dma-controller: external firmware not found, using ROM firmware
Add the SOC_IMX8M into the check so the firmware can load on i.MX8.
Fixes: 1474d48bd639 ("arm64: dts: imx8mq: Add SDMA nodes") Fixes: 941acd566b18 ("dmaengine: imx-sdma: Only check ratio on parts that support 1:1") Signed-off-by: Peter Robinson <pbrobinson@gmail.com> Cc: stable@vger.kernel.org # v5.2+ Reviewed-by: Fabio Estevam <festevam@gmail.com> Link: https://lore.kernel.org/r/20220606161034.3544803-1-pbrobinson@gmail.com Signed-off-by: Vinod Koul <vkoul@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Offloading police with action TC_ACT_UNSPEC was erroneously disabled even
though it was supported by mlx5 matchall offload implementation, which
didn't verify the action type but instead assumed that any single police
action attached to matchall classifier is a 'continue' action. Lack of
action type check made it non-obvious what mlx5 matchall implementation
actually supports and caused implementers and reviewers of referenced
commits to disallow it as a part of improved validation code.
Fixes: b8cd5831c61c ("net: flow_offload: add tc police action parameters") Fixes: b50e462bc22d ("net/sched: act_police: Add extack messages for offload failure") Signed-off-by: Vlad Buslov <vladbu@nvidia.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
# tc filter add dev dummy0 ingress pref 1 proto all matchall skip_sw action police rate 100Mbit burst 10000
Error: cls_matchall: Failed to setup flow action.
We have an error talking to the kernel
# cat /sys/kernel/tracing/trace_pipe
tc-182 [000] b..1. 21.592969: netlink_extack: msg=act_police: Offload not supported when conform/exceed action is "reclassify"
tc-182 [000] ..... 21.592982: netlink_extack: msg=cls_matchall: Failed to setup flow action
# tc filter add dev dummy0 ingress pref 1 proto all matchall skip_sw action police rate 100Mbit burst 10000 conform-exceed drop/continue
Error: cls_matchall: Failed to setup flow action.
We have an error talking to the kernel
# cat /sys/kernel/tracing/trace_pipe
tc-184 [000] b..1. 38.882579: netlink_extack: msg=act_police: Offload not supported when conform/exceed action is "continue"
tc-184 [000] ..... 38.882593: netlink_extack: msg=cls_matchall: Failed to setup flow action
Signed-off-by: Ido Schimmel <idosch@nvidia.com> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The callback is used by various actions to populate the flow action
structure prior to offload. Pass extack to this callback so that the
various actions will be able to report accurate error messages to user
space.
Signed-off-by: Ido Schimmel <idosch@nvidia.com> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
66e4c8d95008 ("net: warn if transport header was not set") added
a check that triggers a warning in r8169, see [0].
The commit referenced in the Fixes tag refers to the change from
which the patch applies cleanly, there's nothing wrong with this
commit. It seems the actual issue (not bug, because the warning
is harmless here) was introduced with bdfa4ed68187
("r8169: use Giant Send").
In mptcp_pm_nl_rm_addr_or_subflow() we always mark as available
the id corresponding to the just removed address.
The used bitmap actually tracks only the local IDs: we must
restrict the operation when a (local) subflow is removed.
Fixes: a88c9e496937 ("mptcp: do not block subflows creation on errors") Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
When setting up a subflow's flags for sending MP_PRIO MPTCP options, the
subflow socket lock was not held while reading and modifying several
struct members that are also read and modified in mptcp_write_options().
Acquire the subflow socket lock earlier and send the MP_PRIO ACK with
that lock already acquired. Add a new variant of the
mptcp_subflow_send_ack() helper to use with the subflow lock held.
Fixes: 067065422fcd ("mptcp: add the outgoing MP_PRIO support") Acked-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The in-kernel path manager code for changing subflow flags acquired both
the msk socket lock and the PM lock when possibly changing the "backup"
and "fullmesh" flags. mptcp_pm_nl_mp_prio_send_ack() does not access
anything protected by the PM lock, and it must release and reacquire
the PM lock.
By pushing the PM lock to where it is needed in mptcp_pm_nl_fullmesh(),
the lock is only acquired when the fullmesh flag is changed and the
backup flag code no longer has to release and reacquire the PM lock. The
change in locking context requires the MIB update to be modified - move
that to a better location instead.
This change also makes it possible to call
mptcp_pm_nl_mp_prio_send_ack() for the userspace PM commands without
manipulating the in-kernel PM lock.
Fixes: 0f9f696a502e ("mptcp: add set_flags command in PM netlink") Acked-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Referenced commit prepared the code for upcoming extension that allows mlx5
to offload police action attached to flower classifier. However, with
regard to existing matchall classifier offload validation should be
reversed as FLOW_ACTION_CONTINUE is the only supported notexceed police
action type. Fix the problem by allowing FLOW_ACTION_CONTINUE for police
action and extend scan_tc_matchall_fdb_actions() to only allow such actions
with matchall classifier.
Fixes: d97b4b105ce7 ("flow_offload: reject offload for all drivers with invalid police parameters") Signed-off-by: Vlad Buslov <vladbu@nvidia.com> Acked-by: Saeed Mahameed <saeedm@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 72f2ecb7ece7 ("ACPI: bus: Set CPPC _OSC bits for all and
when CPPC_LIB is supported") added support for claiming to
support CPPC in _OSC on non-Intel platforms.
This unfortunately caused a regression on a vartiety of AMD
platforms in the field because a number of AMD platforms don't set
the `_OSC` bit 5 or 6 to indicate CPPC or CPPC v2 support.
As these AMD platforms already claim CPPC support via a dedicated
MSR from `X86_FEATURE_CPPC`, use this enable this feature rather
than requiring the `_OSC` on platforms with a dedicated MSR.
If there is additional breakage on the shared memory designs also
missing this _OSC, additional follow up changes may be needed.
Fixes: 72f2ecb7ece7 ("Set CPPC _OSC bits for all and when CPPC_LIB is supported") Reported-by: Perry Yuan <perry.yuan@amd.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Previously the kernel used to ignore whether the firmware masked CPPC
or CPPCv2 and would just pretend that it worked.
When support for the USB4 bit in _OSC was introduced from commit 9e1f561afb ("ACPI: Execute platform _OSC also with query bit clear")
the kernel began to look at the return when the query bit was clear.
This caused regressions that were misdiagnosed and attempted to be solved
as part of commit 2ca8e6285250 ("Revert "ACPI: Pass the same capabilities
to the _OSC regardless of the query flag""). This caused a different
regression where non-Intel systems weren't able to negotiate _OSC
properly.
This was reverted in commit 2ca8e6285250 ("Revert "ACPI: Pass the same
capabilities to the _OSC regardless of the query flag"") and attempted to
be fixed by commit c42fa24b4475 ("ACPI: bus: Avoid using CPPC if not
supported by firmware") but the regression still returned.
These systems with the regression only load support for CPPC from an SSDT
dynamically when _OSC reports CPPC v2. Avoid the problem by not letting
CPPC satisfy the requirement in `acpi_cppc_processor_probe`.
Reported-by: CUI Hao <cuihao.leo@gmail.com> Reported-by: maxim.novozhilov@gmail.com Reported-by: lethe.tree@protonmail.com Reported-by: garystephenwright@gmail.com Reported-by: galaxyking0419@gmail.com Fixes: c42fa24b4475 ("ACPI: bus: Avoid using CPPC if not supported by firmware") Fixes: 2ca8e6285250 ("Revert "ACPI Pass the same capabilities to the _OSC regardless of the query flag"") Link: https://bugzilla.kernel.org/show_bug.cgi?id=213023 Link: https://bugzilla.redhat.com/show_bug.cgi?id=2075387 Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com> Tested-by: CUI Hao <cuihao.leo@gmail.com> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The _OSC method allows the OS and firmware to communicate about
supported features/capabitlities. It also allows the OS to take
control of some features.
In ACPI 6.4, s6.2.11.2 Platform-Wide OSPM Capabilities, the CPPC
(resp. v2) bit should be set by the OS if it 'supports controlling
processor performance via the interfaces described in the _CPC
object'.
The OS supports CPPC and parses the _CPC object only if
CONFIG_ACPI_CPPC_LIB is set. Replace the x86 specific
boot_cpu_has(X86_FEATURE_HWP) dynamic check with an arch
generic CONFIG_ACPI_CPPC_LIB build-time check.
ACPI 6.2 Section 6.2.11.2 'Platform-Wide OSPM Capabilities':
Starting with ACPI Specification 6.2, all _CPC registers can be in
PCC, System Memory, System IO, or Functional Fixed Hardware address
spaces. OSPM support for this more flexible register space scheme is
indicated by the “Flexible Address Space for CPPC Registers” _OSC bit
Otherwise (cf ACPI 6.1, s8.4.7.1.1.X), _CPC registers must be in:
- PCC or Functional Fixed Hardware address space if defined
- SystemMemory address space (NULL register) if not defined
Add the corresponding _OSC bit and check it when parsing _CPC objects.
Signed-off-by: Pierre Gondois <pierre.gondois@arm.com> Reviewed-by: Sudeep Holla <sudeep.holla@arm.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When packets are not received, they aren't received on $host1_if, so the
message talking about the second host not receiving them is incorrect.
Fix it.
Fixes: d4deb01467ec ("selftests: forwarding: Add a test for FDB learning") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The first host interface has by default no interest in receiving packets
MAC DA de:ad:be:ef:13:37, so it might drop them before they hit the tc
filter and this might confuse the selftest.
Enable promiscuous mode such that the filter properly counts received
packets.
Fixes: d4deb01467ec ("selftests: forwarding: Add a test for FDB learning") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
As mentioned in the blamed commit, flood_unicast_test() works by
checking the match count on a tc filter placed on the receiving
interface.
But the second host interface (host2_if) has no interest in receiving a
packet with MAC DA de:ad:be:ef:13:37, so its RX filter drops it even
before the ingress tc filter gets to be executed. So we will incorrectly
get the message "Packet was not flooded when should", when in fact, the
packet was flooded as expected but dropped due to an unrelated reason,
at some other layer on the receiving side.
Force h2 to accept this packet by temporarily placing it in promiscuous
mode. Alternatively we could either deliver to its MAC address or use
tcpdump_start, but this has the fewest complications.
This fixes the "flooding" test from bridge_vlan_aware.sh and
bridge_vlan_unaware.sh, which calls flood_test from the lib.
Fixes: 236dd50bf67a ("selftests: forwarding: Add a test for flooded traffic") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
During a reset, there may have been transmits in flight that are no
longer valid and cannot be fulfilled. Resetting and clearing the
queues is insufficient; each skb also needs to be explicitly freed
so that upper levels are not left waiting for confirmation of a
transmit that will never happen. If this happens frequently enough,
the apparent backlog will cause TCP to begin "congestion control"
unnecessarily, culminating in permanently decreased throughput.
Fixes: d7c0ef36bde03 ("ibmvnic: Free and re-allocate scrqs when tx/rx scrqs change") Tested-by: Nick Child <nnac123@linux.ibm.com> Reviewed-by: Brian King <brking@linux.vnet.ibm.com> Signed-off-by: Rick Lindsley <ricklind@us.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
The USBH composed of EHCI and OHCI controllers needs the PHY clock to be
initialized first, before enabling (gating) them. The reverse is also
required when going to suspend.
So, add USBPHY clock as 1st entry in both controllers, so the USBPHY PLL
gets enabled 1st upon controller init. Upon suspend/resume, this also makes
the clock to be disabled/re-enabled in the correct order.
This fixes some IRQ storm conditions seen when going to low-power, due to
PHY PLL being disabled before all clocks are cleanly gated.
Fixes: 949a0c0dec85 ("ARM: dts: stm32: add USB Host (USBH) support to stm32mp157c") Fixes: db7be2cb87ae ("ARM: dts: stm32: use usbphyc ck_usbo_48m as USBH OHCI clock on stm32mp151") Signed-off-by: Fabrice Gasnier <fabrice.gasnier@foss.st.com> Signed-off-by: Alexandre Torgue <alexandre.torgue@foss.st.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Clear VF MAC from parent PF and remove VF filter from VSI when both
conditions are true:
-VIRTCHNL_VF_OFFLOAD_USO is not used
-VM MAC was not set from PF level
It affects older version of IAVF and it allow them to change MAC
Address on VM, newer IAVF won't change their behaviour.
Previously it wasn't possible to change VF's MAC Address on VM
because there is flag on IAVF driver that won't allow to
change MAC Address if this address is given from PF driver.
Fixes: 155f0ac2c96b ("iavf: allow permanent MAC address to change") Signed-off-by: Norbert Zulinski <norbertx.zulinski@intel.com> Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com> Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Dropped packets caused by too large frames were not included in
dropped RX packets statistics.
Issue was caused by not reading the GL_RXERR1 register. That register
stores count of packet which was have been dropped due to too large
size.
Fix it by reading GL_RXERR1 register for each interface.
Repro steps:
Send a packet larger than the set MTU to SUT
Observe rx statists: ethtool -S <interface> | grep rx | grep -v ": 0"
Fixes: 41a9e55c89be ("i40e: add missing VSI statistics") Signed-off-by: Lukasz Cieplicki <lukaszx.cieplicki@intel.com> Signed-off-by: Jedrzej Jagielski <jedrzej.jagielski@intel.com> Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The recently added support for EFCH MMIO regions introduced a memory
leak in that code path. The leak is caused by the fact that
release_resource() merely removes the resource from the tree but does
not free its memory. We need to call release_mem_region() instead,
which does free the memory. As a nice side effect, this brings back
some symmetry between the legacy and MMIO paths.
Signed-off-by: Jean Delvare <jdelvare@suse.de> Reported-by: Yi Zhang <yi.zhang@redhat.com> Tested-by: Yi Zhang <yi.zhang@redhat.com> Reviewed-by: Terry Bowman <terry.bowman@amd.com> Tested-by: Terry Bowman <Terry.Bowman@amd.com> Fixes: 7c148722d074 ("i2c: piix4: Add EFCH MMIO support to region request and release") Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a XSK pool gets mapped, xp_check_dma_contiguity() adds bit 0x1
to pages' DMA addresses that go in ascending order and at 4K stride.
The problem is that the bit does not get cleared before doing unmap.
As a result, a lot of warnings from iommu_dma_unmap_page() are seen
in dmesg, which indicates that lookups by iommu_iova_to_phys() fail.
Fixes: 2b43470add8c ("xsk: Introduce AF_XDP buffer allocation API") Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: Magnus Karlsson <magnus.karlsson@intel.com> Link: https://lore.kernel.org/bpf/20220628091848.534803-1-ivan.malov@oktetlabs.ru Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix SoC detection for SAM9X60 SiPs:
SAM9X60D5M
SAM9X60D1G
SAM9X60D6K
Fixes: af3a10513cd6 ("drivers: soc: atmel: add per soc id and version match masks") Signed-off-by: Mihai Sain <mihai.sain@microchip.com> Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com> Link: https://lore.kernel.org/r/20220616081344.1978664-1-claudiu.beznea@microchip.com Signed-off-by: Sasha Levin <sashal@kernel.org>
The board has a microchip 24aa025e48 eeprom, which is a 2 Kbits memory,
so it's compatible with at24c02 not at24c32.
Also the size property is wrong, it's not 128 bytes, but 256 bytes.
Thus removing and leaving it to the default (256).
Fixes: 1e5f532c27371 ("ARM: dts: at91: sam9x60: add device tree for soc and board") Signed-off-by: Eugen Hristev <eugen.hristev@microchip.com> Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com> Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com> Link: https://lore.kernel.org/r/20220607090455.80433-1-eugen.hristev@microchip.com Signed-off-by: Sasha Levin <sashal@kernel.org>
"make dtbs_check" complains about the missing "-supply" suffix for
vdd_lvs1_2 which is clearly a typo, originally introduced in the
msm8994-smd-rpm.dtsi file and apparently later copied to
msm8992-xiaomi-libra.dts:
msm8992-lg-bullhead-rev-10/101.dtb: pm8994-regulators: 'vdd_lvs1_2'
does not match any of the regexes:
'.*-supply$', '^((s|l|lvs|5vs)[0-9]*)|(boost-bypass)|(bob)$', 'pinctrl-[0-9]+'
From schema: regulator/qcom,smd-rpm-regulator.yaml
msm8992-xiaomi-libra.dtb: pm8994-regulators: 'vdd_lvs1_2'
does not match any of the regexes:
'.*-supply$', '^((s|l|lvs|5vs)[0-9]*)|(boost-bypass)|(bob)$', 'pinctrl-[0-9]+'
From schema: regulator/qcom,smd-rpm-regulator.yaml
Reported-by: Rob Herring <robh@kernel.org> Cc: Konrad Dybcio <konrad.dybcio@somainline.org> Fixes: f3b2c99e73be ("arm64: dts: Enable onboard SDHCI on msm8992") Fixes: 0f5cdb31e850 ("arm64: dts: qcom: Add Xiaomi Libra (Mi 4C) device tree") Signed-off-by: Stephan Gerhold <stephan.gerhold@kernkonzept.com> Reviewed-by: Konrad Dybcio <konrad.dybcio@somainline.org> Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org> Link: https://lore.kernel.org/r/20220627135938.2901871-1-stephan.gerhold@kernkonzept.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Some Allwinner SoCs have 2 pinctrls (PIO and R_PIO).
Previous implementation used absolute pin numbering and it was incorrect
for R_PIO pinctrl.
It's necessary to take into account the base pin number.
Fixes: 90be64e27621 ("pinctrl: sunxi: implement pin_config_set") Signed-off-by: Andrei Lalaev <andrey.lalaev@gmail.com> Reviewed-by: Samuel Holland <samuel@sholland.org> Link: https://lore.kernel.org/r/20220525190423.410609-1-andrey.lalaev@gmail.com Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
According to the IOMUXC_SW_PAD_CTL_PAD_UART2_RXD/TXD register define in
imx8mp RM, bit0 and bit3 are reserved, and the uart2 rx/tx pin should
enable the pull up, so need to set bit8 to 1. The original pinctl value
0x49 is incorrect and needs to be changed to 0x140, same as uart1 and
uart3.
Currently, when booting Linux on a imx28-evk board there is
no display activity.
Enable CONFIG_FB which is nowadays required for CONFIG_DRM_PANEL_LVDS,
CONFIG_DRM_PANEL_SIMPLE, CONFIG_DRM_PANEL_SEIKO_43WVF1G,
CONFIG_FB_MODE_HELPERS, CONFIG_BACKLIGHT_PWM, CONFIG_BACKLIGHT_GPIO,
CONFIG_FRAMEBUFFER_CONSOLE, CONFIG_LOGO, CONFIG_FONTS, CONFIG_FONT_8x8
and CONFIG_FONT_8x16.
Based on commit c54467482ffd ("ARM: imx_v6_v7_defconfig: enable fb").
It was noticed that on sdm845 after an MDSS suspend/resume cycle the
driver can not read HW_REV registers properly (they will return 0
instead). Chaning the "iface" clock from <&gcc GCC_DISP_AHB_CLK> to
<&dispcc DISP_CC_MDSS_AHB_CLK> fixes the issue.
Commit 288fad2f71fa ("ASoC: SOF: Intel: hda: add quirks for HDAudio DMA position information")
modified the PCM path only, but left the compressed data patch using an
obsolete option.
Move the functionality in a helper that can be called for both PCM and
compressed data.
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com> Fixes: 288fad2f71fa ("ASoC: SOF: Intel: hda: add quirks for HDAudio DMA position information") Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Link: https://lore.kernel.org/r/20220616201953.130876-1-pierre-louis.bossart@linux.intel.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Move the size checks prior to allocating memory as these checks do not need
the data to be allocated and in case of an error we would not need to free
the allocation.
The max size must not be less than the size of
struct sof_ipc_ctrl_data + struct sof_abi_hdr as the ABI header needs to
be present under all circumstances.
The check was incorrectly used or between the two size checks.
Fixes: b5cee8feb1d4 ("ASoC: SOF: topology: Make control parsing IPC agnostic") Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com> Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com> Link: https://lore.kernel.org/r/20220610084735.19397-1-peter.ujfalusi@linux.intel.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The .set_jack_detect() codec component callback is invoked during card
registration, which happens when the machine driver is probed.
The issue is that this callback can race with the bus suspend/resume,
and IO timeouts can happen. This can be reproduced very easily if the
machine driver is 'blacklisted' and manually probed after the bus
suspends. The bus and codec need to be re-initialized using pm_runtime
helpers.
Previous contributions tried to make sure accesses to the bus during
the .set_jack_detect() component callback only happen when the bus is
active. This was done by changing the regcache status on a component
remove. This is however a layering violation, the regcache status
should only be modified on device probe, suspend and resume. The
component probe/remove should not modify how the device regcache is
handled. This solution also didn't handle all the possible race
conditions, and the RT700 headset codec was not handled.
This patch tries to resume the codec device before handling the jack
initializations. In case the codec has not yet been initialized,
pm_runtime may not be enabled yet, so we don't squelch the -EACCES
error code and only stop the jack information. When the codec reports
as attached, the jack initialization will proceed as usual.
The endianness flag is used on the CODEC side to specify an
ambivalence to endian, typically because it is lost over the hardware
link. This device receives audio over a SoundWire DAI and as such
should have endianness applied.
The endianness flag is used on the CODEC side to specify an
ambivalence to endian, typically because it is lost over the hardware
link. This device receives audio over a SoundWire DAI and as such
should have endianness applied.
Currently, cleanup_srcu_struct() checks for a grace period in progress,
but it does not check for a grace period that has not yet started but
which might start at any time. Such a situation could result in a
use-after-free bug, so this commit adds a check for a grace period that
is needed but not yet started to cleanup_srcu_struct().
Signed-off-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
of_find_compatible_node() returns a node pointer with refcount
incremented, we should use of_node_put() on it when done.
Add missing of_node_put() to avoid refcount leak.
Fixes: d850f3e5d296 ("ARM: meson: Add SMP bringup code for Meson8 and Meson8b") Signed-off-by: Miaoqian Lin <linmq006@gmail.com> Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Link: https://lore.kernel.org/r/20220512021611.47921-1-linmq006@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
It was discovered that the Documentation lacks of a fundamental detail
on how to correctly change the MAX_FRAME_SIZE of the switch.
In fact if the MAX_FRAME_SIZE is changed while the cpu port is on, the
switch panics and cease to send any packet. This cause the mgmt ethernet
system to not receive any packet (the slow fallback still works) and
makes the device not reachable. To recover from this a switch reset is
required.
To correctly handle this, turn off the cpu ports before changing the
MAX_FRAME_SIZE and turn on again after the value is applied.
Fixes: f58d2598cf70 ("net: dsa: qca8k: implement the port MTU callbacks") Cc: stable@vger.kernel.org Signed-off-by: Christian Marangi <ansuelsmth@gmail.com> Link: https://lore.kernel.org/r/20220621151122.10220-1-ansuelsmth@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The platform device for the rng must be created much later in boot.
Otherwise it tries to connect to a parent that doesn't yet exist,
resulting in this splat:
This patch fixes the issue by doing the platform device creation inside
of machine_subsys_initcall.
Fixes: f3eac426657d ("powerpc/powernv: wire up rng during setup_arch") Cc: stable@vger.kernel.org Reported-by: Sachin Sant <sachinp@linux.ibm.com> Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com> Tested-by: Sachin Sant <sachinp@linux.ibm.com>
[mpe: Change "of node" to "platform device" in change log] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20220630121654.1939181-1-Jason@zx2c4.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After waiting for the volume to complete the acquisition with timeout,
the if condition under which potential volume collision occurs should be
acquire the volume is still pending rather than not pending so that we
will continue to wait until the pending flag is cleared. Also, use the
existing test pending wrapper directly instead of test_bit().
If an NFS file is opened for writing and closed, fscache_invalidate() will
be asked to invalidate the file - however, if the cookie is in the
LOOKING_UP state (or the CREATING state), then request to invalidate
doesn't get recorded for fscache_cookie_state_machine() to do something
with.
Fix this by making __fscache_invalidate() set a flag if it sees the cookie
is in the LOOKING_UP state to indicate that we need to go to invalidation.
Note that this requires a count on the n_accesses counter for the state
machine, which that will release when it's done.
fscache_cookie_state_machine() then shifts to the INVALIDATING state if it
sees the flag.
Without this, an nfs file can get corrupted if it gets modified locally and
then read locally as the cache contents may not get updated.
Fixes: d24af13e2e23 ("fscache: Implement cookie invalidation") Reported-by: Max Kellermann <mk@cm4all.com> Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Max Kellermann <mk@cm4all.com> Link: https://lore.kernel.org/r/YlWWbpW5Foynjllo@rabbit.intern.cm-ag Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the CONFIG_MEMREGION=n case, memregion_free() is meant to be a static
inline. 0day reports:
In file included from drivers/cxl/core/port.c:4:
include/linux/memregion.h:19:6: warning: no previous prototype for
function 'memregion_free' [-Wmissing-prototypes]
Mark memregion_free() static.
Fixes: 33dd70752cd7 ("lib: Uplevel the pmem "region" ida to a global allocator") Reported-by: kernel test robot <lkp@intel.com> Reviewed-by: Alison Schofield <alison.schofield@intel.com> Link: https://lore.kernel.org/r/165601455171.4042645.3350844271068713515.stgit@dwillia2-xfh Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Because pm_runtime_get_suppliers() bumps up the rpm_active counter
of each device link to a supplier of the given device in addition
to bumping up the supplier's PM-runtime usage counter, a runtime
suspend of the consumer device may case the latter to go down to 0
when pm_runtime_put_suppliers() is running on a remote CPU. If that
happens after pm_runtime_put_suppliers() has released power.lock for
the consumer device, and a runtime resume of that device takes place
immediately after it, before pm_runtime_put() is called for the
supplier, that pm_runtime_put() call may cause the supplier to be
suspended even though the consumer is active.
To prevent that from happening, modify pm_runtime_get_suppliers() to
call pm_runtime_get_sync() for the given device's suppliers without
touching the rpm_active counters of the involved device links
Accordingly, modify pm_runtime_put_suppliers() to call pm_runtime_put()
for the given device's suppliers without looking at the rpm_active
counters of the device links at hand. [This is analogous to what
happened before commit 4c06c4e6cf63 ("driver core: Fix possible
supplier PM-usage counter imbalance").]
Since pm_runtime_get_suppliers() sets supplier_preactivated for each
device link where the supplier's PM-runtime usage counter has been
incremented and pm_runtime_put_suppliers() calls pm_runtime_put() for
the suppliers whose device links have supplier_preactivated set, the
PM-runtime usage counter is balanced for each supplier and this is
independent of the runtime suspend and resume of the consumer device.
However, in case a device link with DL_FLAG_PM_RUNTIME set is dropped
during the consumer device probe, so pm_runtime_get_suppliers() bumps
up the supplier's PM-runtime usage counter, but it cannot be dropped by
pm_runtime_put_suppliers(), make device_link_release_fn() take care of
that.
Fixes: 4c06c4e6cf63 ("driver core: Fix possible supplier PM-usage counter imbalance") Reported-by: Peter Wang <peter.wang@mediatek.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Peter Wang <peter.wang@mediatek.com> Cc: 5.1+ <stable@vger.kernel.org> # 5.1+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Instead of passing an extra bool argument to pm_runtime_release_supplier(),
make its callers take care of triggering a runtime-suspend of the
supplier device as needed.
We need to prevent that users configure a screen size which is smaller than the
currently selected font size. Otherwise rendering chars on the screen will
access memory outside the graphics memory region.
This patch adds a new function fbcon_modechange_possible() which
implements this check and which later may be extended with other checks
if necessary. The new function is called from the FBIOPUT_VSCREENINFO
ioctl handler in fbmem.c, which will return -EINVAL if userspace asked
for a too small screen size.
Prevent that users set a font size which is bigger than the physical screen.
It's unlikely this may happen (because screens are usually much larger than the
fonts and each font char is limited to 32x32 pixels), but it may happen on
smaller screens/LCD displays.
Verify that the fbdev or drm driver correctly adjusted the virtual
screen sizes. On failure report the failing driver and reject the screen
size change.
The device is created, and then there is a check if a driver succesfully
bound to it. In event of failing the bind (e.g. failure in cxl_port_probe())
the device is left registered. When a bus rescan later occurs, fresh
devices are created leading to a multiple device representing the same
underlying hardware. Bad things may follow and at very least we have far too many
devices.
Fix by ensuring autoremove is registered if the device create succeeds,
but doesn't depend on sucessful binding to a driver.
Bug was observed as side effect of incorrect ownership in
[PATCH v9 6/9] cxl/port: Read CDAT table
but will result from any failure to in cxl_port_probe().
Fixes: 8dd2bc0f8e02 ("cxl/mem: Add the cxl_mem driver") Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Link: https://lore.kernel.org/r/20220609134519.11668-1-Jonathan.Cameron@huawei.com Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The IOMMU driver shares the pasid table for PCI alias devices. When the
RID2PASID entry of the shared pasid table has been filled by the first
device, the subsequent device will encounter the "DMAR: Setup RID2PASID
failed" failure as the pasid entry has already been marked as present.
As the result, the IOMMU probing process will be aborted.
On the contrary, when any alias device is hot-removed from the system,
for example, by writing to /sys/bus/pci/devices/.../remove, the shared
RID2PASID will be cleared without any notifications to other devices.
As the result, any DMAs from those rest devices are blocked.
Sharing pasid table among PCI alias devices could save two memory pages
for devices underneath the PCIe-to-PCI bridges. Anyway, considering that
those devices are rare on modern platforms that support VT-d in scalable
mode and the saved memory is negligible, it's reasonable to remove this
part of immature code to make the driver feasible and stable.
Notifier calling chain uses priority to determine the execution
order of the notifiers or listeners registered to the chain.
PCI bus device hot add utilizes the notification mechanism.
The current code sets low priority (INT_MIN) to Intel
dmar_pci_bus_notifier and postpones DMAR decoding after adding
new device into IOMMU. The result is that struct device pointer
cannot be found in DRHD search for the new device's DMAR/IOMMU.
Subsequently, the device is put under the "catch-all" IOMMU
instead of the correct one. This could cause system hang when
device TLB invalidation is sent to the wrong IOMMU. Invalidation
timeout error and hard lockup have been observed and data
inconsistency/crush may occur as well.
This patch fixes the issue by setting a positive priority(1) for
dmar_pci_bus_notifier while the priority of IOMMU bus notifier
uses the default value(0), therefore DMAR decoding will be in
advance of DRHD search for a new device to find the correct IOMMU.
Following is a 2-step example that triggers the bug by simulating
PCI device hot add behavior in Intel Sapphire Rapids server.
In mcp251xfd_register_get_dev_id() the device ID register is read with
handcrafted SPI transfers. As all registers, this register is in
little endian. Further it is not naturally aligned in struct
mcp251xfd_map_buf_nocrc::data. However after the transfer the register
content is converted from big endian to CPU endianness not taking care
of being unaligned.
Fix the conversion by converting from little endian to CPU endianness
taking the unaligned source into account.
Side note: So far the register content is 0x0 on all mcp251xfd
compatible chips, and is only used for an informative printk.
The device ID register is 32 bits wide. The driver uses incorrectly
the size of a pointer to a u32 to calculate the length of the SPI
transfer. This results in a read of 2 registers on 64 bit platforms.
This is no problem on the Linux side, as the RX buffer of the SPI
transfer is large enough. In the mpc251xfd chip this results in the
read of an undocumented register. So far no problems were observed.
Fix the length of the SPI transfer to read the device ID register
only.
Since commit 8fffa0e3451a ("selftests/bpf: Normalize XDP section names in
selftests") the xdp_dummy.o's section name has changed to xdp. But some
tests are still using "section xdp_dummy", which make the tests failed.
Fix them by updating to the new section name.
Fixes: 8fffa0e3451a ("selftests/bpf: Normalize XDP section names in selftests") Signed-off-by: Hangbin Liu <liuhangbin@gmail.com> Acked-by: Andrii Nakryiko <andrii@kernel.org> Link: https://lore.kernel.org/r/20220630062228.3453016-1-liuhangbin@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add nft_set_pipapo_match_destroy() helper function to release the
elements in the lookup tables.
Stefano Brivio says: "We additionally look for elements pointers in the
cloned matching data if priv->dirty is set, because that means that
cloned data might point to additional elements we did not commit to the
working copy yet (such as the abort path case, but perhaps not limited
to it)."
Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges") Reviewed-by: Stefano Brivio <sbrivio@redhat.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Instead of counting the child nodes in the device tree, hardcode the
number of ports in the driver itself. The counting won't work at all
if an ethernet port is marked as disabled, e.g. because it is not
connected on the board at all.
It turns out that the LAN9662 and LAN9668 use the same switching IP
with the same synthesis parameters. The only difference is that the
output ports are not connected. Thus, we can just hardcode the
number of physical ports to 8.
Fixes: db8bcaad5393 ("net: lan966x: add the basic lan966x driver") Signed-off-by: Michael Walle <michael@walle.cc> Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com> Link: https://lore.kernel.org/r/20220704153654.1167886-1-michael@walle.cc Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There are UAF bugs caused by rose_t0timer_expiry(). The
root cause is that del_timer() could not stop the timer
handler that is running and there is no synchronization.
One of the race conditions is shown below:
The rose_neigh is deallocated in position [1] and use in
position [2].
The crash trace triggered by POC is like below:
BUG: KASAN: use-after-free in expire_timers+0x144/0x320
Write of size 8 at addr ffff888009b19658 by task swapper/0/0
...
Call Trace:
<IRQ>
dump_stack_lvl+0xbf/0xee
print_address_description+0x7b/0x440
print_report+0x101/0x230
? expire_timers+0x144/0x320
kasan_report+0xed/0x120
? expire_timers+0x144/0x320
expire_timers+0x144/0x320
__run_timers+0x3ff/0x4d0
run_timer_softirq+0x41/0x80
__do_softirq+0x233/0x544
...
This patch changes rose_stop_ftimer() and rose_stop_t0timer()
in rose_remove_neigh() to del_timer_sync() in order that the
timer handler could be finished before the resources such as
rose_neigh and so on are deallocated. As a result, the UAF
bugs could be mitigated.
Kuee reported a corner case where the tnum becomes constant after the call
to __reg_bound_offset(), but the register's bounds are not, that is, its
min bounds are still not equal to the register's max bounds.
This in turn allows to leak pointers through turning a pointer register as
is into an unknown scalar via adjust_ptr_min_max_vals().
What can be seen here is that R3=scalar(umin=32767,umax=32768,var_off=(0x7fff;
0x8000)) after the operation R3 += -32767 results in a 'malformed' constant, that
is, R3_w=scalar(imm=0,umax=1,var_off=(0x0; 0x0)). Intersecting with var_off has
not been done at that point via __update_reg_bounds(), which would have improved
the umax to be equal to umin.
Refactor the tnum <> min/max bounds information flow into a reg_bounds_sync()
helper and use it consistently everywhere. After the fix, bounds have been
corrected to R3_w=scalar(imm=0,umax=0,var_off=(0x0; 0x0)) and thus the register
is regarded as a 'proper' constant scalar of 0.
Kuee reported a quirk in the jmp32's jeq/jne simulation, namely that the
register value does not match expectations for the fall-through path. For
example:
As can be seen on line 5 for the branch fall-through path in R2 [*] is that
given condition w2 != 0x8 is false, verifier should conclude that r2 = 8 as
upper 32 bit are known to be zero. However, verifier incorrectly concludes
that r2 = 571 which is far off.
The problem is it only marks false{true}_reg as known in the switch for JE/NE
case, but at the end of the function, it uses {false,true}_{64,32}off to
update {false,true}_reg->var_off and they still hold the prior value of
{false,true}_reg->var_off before it got marked as known. The subsequent
__reg_combine_32_into_64() then propagates this old var_off and derives new
bounds. The information between min/max bounds on {false,true}_reg from
setting the register to known const combined with the {false,true}_reg->var_off
based on the old information then derives wrong register data.
Fix it by detangling the BPF_JEQ/BPF_JNE cases and updating relevant
{false,true}_{64,32}off tnums along with the register marking to known
constant.