Due to the way the GDS and SRSO patches flowed into the stable tree, it
was a 50% chance that the merge of the which value GDS and SRSO should
be. Of course, I lost that bet, and chose the opposite of what Linus
chose in commit 64094e7e3118 ("Merge tag 'gds-for-linus-2023-08-01' of
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")
Fix this up by switching the values to match what is now in Linus's tree
as that is the correct value to mirror.
It is possible that a guest can send a packet that contains a head + 18
slots and yet has a len <= XEN_NETBACK_TX_COPY_LEN. This causes nr_slots
to underflow in xenvif_get_requests() which then causes the subsequent
loop's termination condition to be wrong, causing a buffer overrun of
queue->tx_map_ops.
Rework the code to account for the extra frag_overflow slots.
This is CVE-2023-34319 / XSA-432.
Fixes: ad7f402ae4f4 ("xen/netback: Ensure protocol headers don't fall in the non-linear area") Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com> Reviewed-by: Paul Durrant <paul@xen.org> Reviewed-by: Wei Liu <wei.liu@kernel.org> Signed-off-by: Juergen Gross <jgross@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
3b9c723ed7cfa4e1 ("KVM: SVM: Add support for SVM instruction address check change") b85a0425d8056f3b ("Enumerate AVX Vector Neural Network instructions") fb35d30fe5b06cc2 ("x86/cpufeatures: Assign dedicated feature word for CPUID_0x8000001F[EAX]")
This only causes these perf files to be rebuilt:
CC /tmp/build/perf/bench/mem-memcpy-x86-64-asm.o
CC /tmp/build/perf/bench/mem-memset-x86-64-asm.o
And addresses this perf build warning:
Warning: Kernel ABI header at 'tools/arch/x86/include/asm/cpufeatures.h' differs from latest version at 'arch/x86/include/asm/cpufeatures.h'
diff -u tools/arch/x86/include/asm/cpufeatures.h arch/x86/include/asm/cpufeatures.h
Cc: Borislav Petkov <bp@suse.de> Cc: Kyung Min Park <kyung.min.park@intel.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Sean Christopherson <seanjc@google.com> Cc: Wei Huang <wei.huang2@amd.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Collect the scattered SME/SEV related feature flags into a dedicated
word. There are now five recognized features in CPUID.0x8000001F.EAX,
with at least one more on the horizon (SEV-SNP). Using a dedicated word
allows KVM to use its automagic CPUID adjustment logic when reporting
the set of supported features to userspace.
On systems that do not have hardware enforced cache coherency between
encrypted and unencrypted mappings of the same physical page, the
hypervisor can use the VM page flush MSR (0xc001011e) to flush the cache
contents of an SEV guest page. When a small number of pages are being
flushed, this can be used in place of issuing a WBINVD across all CPUs.
CPUID 0x8000001f_eax[2] is used to determine if the VM page flush MSR is
available. Add a CPUID feature to indicate it is supported and define the
MSR.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Message-Id: <f1966379e31f9b208db5257509c4a089a87d33d0.1607620209.git.thomas.lendacky@amd.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add CPU feature detection for Secure Encrypted Virtualization with
Encrypted State. This feature enhances SEV by also encrypting the
guest register state, making it in-accessible to the hypervisor.
Instead of duplicating init_mm, allocate a fresh mm. The advantage is
that mm_alloc() has much simpler dependencies. Additionally it makes
more conceptual sense, init_mm has no (and must not have) user state
to duplicate.
Commit 3f4c8211d982 ("x86/mm: Use mm_alloc() in poking_init()") broke
the kernel for running as Xen PV guest.
It seems as if the new address space is never activated before being
used, resulting in Xen rejecting to accept the new CR3 value (the PGD
isn't pinned).
Fix that by adding the now missing call of paravirt_arch_dup_mmap() to
poking_init(). That call was previously done by dup_mm()->dup_mmap() and
it is a NOP for all cases but for Xen PV, where it is just doing the
pinning of the PGD.
Fixes: 3f4c8211d982 ("x86/mm: Use mm_alloc() in poking_init()") Signed-off-by: Juergen Gross <jgross@suse.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20230109150922.10578-1-jgross@suse.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Gather Data Sampling (GDS) is a transient execution attack using
gather instructions from the AVX2 and AVX512 extensions. This attack
allows malicious code to infer data that was previously stored in
vector registers. Systems that are not vulnerable to GDS will set the
GDS_NO bit of the IA32_ARCH_CAPABILITIES MSR. This is useful for VM
guests that may think they are on vulnerable systems that are, in
fact, not affected. Guests that are running on affected hosts where
the mitigation is enabled are protected as if they were running
on an unaffected system.
On all hosts that are not affected or that are mitigated, set the
GDS_NO bit.
Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Gather Data Sampling (GDS) is mitigated in microcode. However, on
systems that haven't received the updated microcode, disabling AVX
can act as a mitigation. Add a Kconfig option that uses the microcode
mitigation if available and disables AVX otherwise. Setting this
option has no effect on systems not affected by GDS. This is the
equivalent of setting gather_data_sampling=force.
Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The Gather Data Sampling (GDS) vulnerability allows malicious software
to infer stale data previously stored in vector registers. This may
include sensitive data such as cryptographic keys. GDS is mitigated in
microcode, and systems with up-to-date microcode are protected by
default. However, any affected system that is running with older
microcode will still be vulnerable to GDS attacks.
Since the gather instructions used by the attacker are part of the
AVX2 and AVX512 extensions, disabling these extensions prevents gather
instructions from being executed, thereby mitigating the system from
GDS. Disabling AVX2 is sufficient, but we don't have the granularity
to do this. The XCR0[2] disables AVX, with no option to just disable
AVX2.
Add a kernel parameter gather_data_sampling=force that will enable the
microcode mitigation if available, otherwise it will disable AVX on
affected systems.
This option will be ignored if cmdline mitigations=off.
This is a *big* hammer. It is known to break buggy userspace that
uses incomplete, buggy AVX enumeration. Unfortunately, such userspace
does exist in the wild:
Gather Data Sampling (GDS) is a hardware vulnerability which allows
unprivileged speculative access to data which was previously stored in
vector registers.
Intel processors that support AVX2 and AVX512 have gather instructions
that fetch non-contiguous data elements from memory. On vulnerable
hardware, when a gather instruction is transiently executed and
encounters a fault, stale data from architectural or internal vector
registers may get transiently stored to the destination vector
register allowing an attacker to infer the stale data using typical
side channel techniques like cache timing attacks.
This mitigation is different from many earlier ones for two reasons.
First, it is enabled by default and a bit must be set to *DISABLE* it.
This is the opposite of normal mitigation polarity. This means GDS can
be mitigated simply by updating microcode and leaving the new control
bit alone.
Second, GDS has a "lock" bit. This lock bit is there because the
mitigation affects the hardware security features KeyLocker and SGX.
It needs to be enabled and *STAY* enabled for these features to be
mitigated against GDS.
The mitigation is enabled in the microcode by default. Disable it by
setting gather_data_sampling=off or by disabling all mitigations with
mitigations=off. The mitigation status can be checked by reading:
Initializing the FPU during the early boot process is a pointless
exercise. Early boot is convoluted and fragile enough.
Nothing requires that the FPU is set up early. It has to be initialized
before fork_init() because the task_struct size depends on the FPU register
buffer size.
Move the initialization to arch_cpu_finalize_init() which is the perfect
place to do so.
No functional change.
This allows to remove quite some of the custom early command line parsing,
but that's subject to the next installment.
X86 is reworking the boot process so that initializations which are not
required during early boot can be moved into the late boot process and out
of the fragile and restricted initial boot phase.
arch_cpu_finalize_init() is the obvious place to do such initializations,
but arch_cpu_finalize_init() is invoked too late in start_kernel() e.g. for
initializing the FPU completely. fork_init() requires that the FPU is
initialized as the size of task_struct on X86 depends on the size of the
required FPU register buffer.
Fortunately none of the init calls between calibrate_delay() and
arch_cpu_finalize_init() is relevant for the functionality of
arch_cpu_finalize_init().
Invoke it right after calibrate_delay() where everything which is relevant
for arch_cpu_finalize_init() has been set up already.
No functional change intended.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Rick Edgecombe <rick.p.edgecombe@intel.com> Link: https://lore.kernel.org/r/20230613224545.612182854@linutronix.de Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Everything is converted over to arch_cpu_finalize_init(). Remove the
check_bugs() leftovers including the empty stubs in asm-generic, alpha,
parisc, powerpc and xtensa.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Link: https://lore.kernel.org/r/20230613224545.553215951@linutronix.de Signed-off-by: Daniel Sneddon <daniel.sneddon@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
check_bugs() has become a dumping ground for all sorts of activities to
finalize the CPU initialization before running the rest of the init code.
Most are empty, a few do actual bug checks, some do alternative patching
and some cobble a CPU advertisement string together....
Aside of that the current implementation requires duplicated function
declaration and mostly empty header files for them.
Provide a new function arch_cpu_finalize_init(). Provide a generic
declaration if CONFIG_ARCH_HAS_CPU_FINALIZE_INIT is selected and a stub
inline otherwise.
This requires a temporary #ifdef in start_kernel() which will be removed
along with check_bugs() once the architectures are converted over.
Commit 6018b585e8c6 ("tracing/histograms: Add histograms to hist_vars if
they have referenced variables") added a check to fail histogram creation
if save_hist_vars() failed to add histogram to hist_vars list. But the
commit failed to set ret to failed return code before jumping to
unregister histogram, fix it.
Link: https://lore.kernel.org/linux-trace-kernel/20230714203341.51396-1-mkhalfella@purestorage.com Cc: stable@vger.kernel.org Fixes: 6018b585e8c6 ("tracing/histograms: Add histograms to hist_vars if they have referenced variables") Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes: 1536e2857bd3 ("tcp: Add a TCP_FASTOPEN socket option to get a max backlog on its listner") Signed-off-by: Eric Dumazet <edumazet@google.com> Link: https://lore.kernel.org/r/20230719212857.3943972-12-edumazet@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch changes the behavior of TCP_LINGER2 about its limit. The
sysctl_tcp_fin_timeout used to be the limit of TCP_LINGER2 but now it's
only the default value. A new macro named TCP_FIN_TIMEOUT_MAX is added
as the limit of TCP_LINGER2, which is 2 minutes.
Since TCP_LINGER2 used sysctl_tcp_fin_timeout as the default value
and the limit in the past, the system administrator cannot set the
default value for most of sockets and let some sockets have a greater
timeout. It might be a mistake that let the sysctl to be the limit of
the TCP_LINGER2. Maybe we can add a new sysctl to set the max of
TCP_LINGER2, but FIN-WAIT-2 timeout is usually no need to be too long
and 2 minutes are legal considering TCP specs.
Changes in v3:
- Remove the new socket option and change the TCP_LINGER2 behavior so
that the timeout can be set to value between sysctl_tcp_fin_timeout
and 2 minutes.
Changes in v2:
- Add int overflow check for the new socket option.
Changes in v1:
- Add a new socket option to set timeout greater than
sysctl_tcp_fin_timeout.
Signed-off-by: Cambda Zhu <cambda@linux.alibaba.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Stable-dep-of: 9df5335ca974 ("tcp: annotate data-races around tp->linger2") Signed-off-by: Sasha Levin <sashal@kernel.org>
On some platforms there is a padding hole in the nft_verdict
structure, between the verdict code and the chain pointer.
On element insertion, if the new element clashes with an existing one and
NLM_F_EXCL flag isn't set, we want to ignore the -EEXIST error as long as
the data associated with duplicated element is the same as the existing
one. The data equality check uses memcmp.
For normal data (NFT_DATA_VALUE) this works fine, but for NFT_DATA_VERDICT
padding area leads to spurious failure even if the verdict data is the
same.
This then makes the insertion fail with 'already exists' error, even
though the new "key : data" matches an existing entry and userspace
told the kernel that it doesn't want to receive an error indication.
Fixes: c016c7e45ddf ("netfilter: nf_tables: honor NLM_F_EXCL flag in set element insertion") Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
This func misses checking for platform_get_irq()'s call and may passes the
negative error codes to request_irq(), which takes unsigned IRQ #,
causing it to fail with -EINVAL, overriding an original error code.
Fix this by stop calling request_irq() with invalid IRQ #s.
Commit 3f4ca5fafc08 ("tcp: avoid the lookup process failing to get sk in
ehash table") reversed the order in how a socket is inserted into ehash
to fix an issue that ehash-lookup could fail when reqsk/full sk/twsk are
swapped. However, it introduced another lookup failure.
The full socket in ehash is allocated from a slab with SLAB_TYPESAFE_BY_RCU
and does not have SOCK_RCU_FREE, so the socket could be reused even while
it is being referenced on another CPU doing RCU lookup.
Let's say a socket is reused and inserted into the same hash bucket during
lookup. After the blamed commit, a new socket is inserted at the end of
the list. If that happens, we will skip sockets placed after the previous
position of the reused socket, resulting in ehash lookup failure.
As described in Documentation/RCU/rculist_nulls.rst, we should insert a
new socket at the head of the list to avoid such an issue.
This issue, the swap-lookup-failure, and another variant reported in [0]
can all be handled properly by adding a locked ehash lookup suggested by
Eric Dumazet [1].
However, this issue could occur for every packet, thus more likely than
the other two races, so let's revert the change for now.
We do netif_napi_add() for all allocated q_vectors[], but potentially
do netif_napi_del() for part of them, then kfree q_vectors and leave
invalid pointers at dev->napi_list.
Although the patch #2 (of 2) can avoid the issue triggered by this
repro.sh, there still are other potential risks that if num_active_queues
is changed to less than allocated q_vectors[] by unexpected, the
mismatched netif_napi_add/del() can also cause UAF.
Since we actually call netif_napi_add() for all allocated q_vectors
unconditionally in iavf_alloc_q_vectors(), so we should fix it by
letting netif_napi_del() match to netif_napi_add().
Fixes: 5eae00c57f5e ("i40evf: main driver core") Signed-off-by: Ding Hui <dinghui@sangfor.com.cn> Cc: Donglin Peng <pengdonglin@sangfor.com.cn> Cc: Huang Cun <huangcun@sangfor.com.cn> Reviewed-by: Simon Horman <simon.horman@corigine.com> Reviewed-by: Madhu Chittim <madhu.chittim@intel.com> Reviewed-by: Leon Romanovsky <leonro@nvidia.com> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
CPSW ALE has 75 bit ALE entries which are stored within three 32 bit words.
The cpsw_ale_get_field() and cpsw_ale_set_field() functions assume that the
field will be strictly contained within one word. However, this is not
guaranteed to be the case and it is possible for ALE field entries to span
across up to two words at the most.
Fix the methods to handle getting/setting fields spanning up to two words.
Fixes: db82173f23c5 ("netdev: driver: ethernet: add cpsw address lookup engine support") Signed-off-by: Tanmay Patil <t-patil@ti.com>
[s-vadapalli@ti.com: rephrased commit message and added Fixes tag] Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
On ASUS TUF A16 it is reported that the ITE5570 ACPI device connected to
GPIO 7 is causing an interrupt storm. This issue doesn't happen on
Windows.
Comparing the GPIO register configuration between Windows and Linux
bit 20 has been configured as a pull up on Windows, but not on Linux.
Checking GPIO declaration from the firmware it is clear it *should* have
been a pull up on Linux as well.
On Linux amd_gpio_set_config() is currently only used for programming
the debounce. Actually the GPIO core calls it with all the arguments
that are supported by a GPIO, pinctrl-amd just responds `-ENOTSUPP`.
To solve this issue expand amd_gpio_set_config() to support the other
arguments amd_pinconf_set() supports, namely `PIN_CONFIG_BIAS_PULL_DOWN`,
`PIN_CONFIG_BIAS_PULL_UP`, and `PIN_CONFIG_DRIVE_STRENGTH`.
Reported-by: Nik P <npliashechnikov@gmail.com> Reported-by: Nathan Schulte <nmschulte@gmail.com> Reported-by: Friedrich Vock <friedrich.vock@gmx.de> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217336 Reported-by: dridri85@gmail.com Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217493 Link: https://lore.kernel.org/linux-input/20230530154058.17594-1-friedrich.vock@gmx.de/ Tested-by: Jan Visser <starquake@linuxeverywhere.org> Fixes: 2956b5d94a76 ("pinctrl / gpio: Introduce .set_config() callback for GPIO chips") Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20230705133005.577-3-mario.limonciello@amd.com Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the prepend byte count field starts at bit 8, and the next defined
bit is SPI_CMD_ONE_BYTE at bit 11, it can be at most 3 bits wide, and
thus the max value is 7, not 15.
Fixes: b17de076062a ("spi/bcm63xx: work around inability to keep CS up") Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com> Link: https://lore.kernel.org/r/20230629071453.62024-1-jonas.gorski@gmail.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In a setup where a Thunderbolt hub connects to Ethernet and a display
through USB Type-C, users may experience a hung task timeout when they
remove the cable between the PC and the Thunderbolt hub.
This is because the igb_down function is called multiple times when
the Thunderbolt hub is unplugged. For example, the igb_io_error_detected
triggers the first call, and the igb_remove triggers the second call.
The second call to igb_down will block at napi_synchronize.
Here's the call trace:
__schedule+0x3b0/0xddb
? __mod_timer+0x164/0x5d3
schedule+0x44/0xa8
schedule_timeout+0xb2/0x2a4
? run_local_timers+0x4e/0x4e
msleep+0x31/0x38
igb_down+0x12c/0x22a [igb 6615058754948bfde0bf01429257eb59f13030d4]
__igb_close+0x6f/0x9c [igb 6615058754948bfde0bf01429257eb59f13030d4]
igb_close+0x23/0x2b [igb 6615058754948bfde0bf01429257eb59f13030d4]
__dev_close_many+0x95/0xec
dev_close_many+0x6e/0x103
unregister_netdevice_many+0x105/0x5b1
unregister_netdevice_queue+0xc2/0x10d
unregister_netdev+0x1c/0x23
igb_remove+0xa7/0x11c [igb 6615058754948bfde0bf01429257eb59f13030d4]
pci_device_remove+0x3f/0x9c
device_release_driver_internal+0xfe/0x1b4
pci_stop_bus_device+0x5b/0x7f
pci_stop_bus_device+0x30/0x7f
pci_stop_bus_device+0x30/0x7f
pci_stop_and_remove_bus_device+0x12/0x19
pciehp_unconfigure_device+0x76/0xe9
pciehp_disable_slot+0x6e/0x131
pciehp_handle_presence_or_link_change+0x7a/0x3f7
pciehp_ist+0xbe/0x194
irq_thread_fn+0x22/0x4d
? irq_thread+0x1fd/0x1fd
irq_thread+0x17b/0x1fd
? irq_forced_thread_fn+0x5f/0x5f
kthread+0x142/0x153
? __irq_get_irqchip_state+0x46/0x46
? kthread_associate_blkcg+0x71/0x71
ret_from_fork+0x1f/0x30
In this case, igb_io_error_detected detaches the network interface
and requests a PCIE slot reset, however, the PCIE reset callback is
not being invoked and thus the Ethernet connection breaks down.
As the PCIE error in this case is a non-fatal one, requesting a
slot reset can be avoided.
This patch fixes the task hung issue and preserves Ethernet
connection by ignoring non-fatal PCIE errors.
Signed-off-by: Ying Hsu <yinghsu@chromium.org> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Reviewed-by: Simon Horman <simon.horman@corigine.com> Link: https://lore.kernel.org/r/20230620174732.4145155-1-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Roee reported various hard-to-debug crashes with pings in
EHT aggregation scenarios. Enabling KASAN showed that we
access the BAID allocation out of bounds, and looking at
the code a bit shows that since the reorder buffer entry
(struct iwl_mvm_reorder_buf_entry) is 128 bytes if debug
such as lockdep is enabled, then staring from an agg size
512 we overflow the size calculation, and allocate a much
smaller structure than we should, causing slab corruption
once we initialize this.
-Wstringop-overflow is legitimately warning us about extra_size
pontentially being zero at some point, hence potenially ending
up _allocating_ zero bytes of memory for extra pointer and then
trying to access such object in a call to copy_from_user().
Fix this by adding a sanity check to ensure we never end up
trying to allocate zero bytes of data for extra pointer, before
continue executing the rest of the code in the function.
Address the following -Wstringop-overflow warning seen when built
m68k architecture with allyesconfig configuration:
from net/wireless/wext-core.c:11:
In function '_copy_from_user',
inlined from 'copy_from_user' at include/linux/uaccess.h:183:7,
inlined from 'ioctl_standard_iw_point' at net/wireless/wext-core.c:825:7:
arch/m68k/include/asm/string.h:48:25: warning: '__builtin_memset' writing 1 or more bytes into a region of size 0 overflows the destination [-Wstringop-overflow=]
48 | #define memset(d, c, n) __builtin_memset(d, c, n)
| ^~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/uaccess.h:153:17: note: in expansion of macro 'memset'
153 | memset(to + (n - res), 0, res);
| ^~~~~~
In function 'kmalloc',
inlined from 'kzalloc' at include/linux/slab.h:694:9,
inlined from 'ioctl_standard_iw_point' at net/wireless/wext-core.c:819:10:
include/linux/slab.h:577:16: note: at offset 1 into destination object of size 0 allocated by '__kmalloc'
577 | return __kmalloc(size, flags);
| ^~~~~~~~~~~~~~~~~~~~~~
This help with the ongoing efforts to globally enable
-Wstringop-overflow.
devlink_port_type_warn is scheduled for port devlink and warning
when the port type is not set. But from this warning it is not easy
found out which device (driver) has no devlink port set.
KCSAN reported a data-race when accessing node->ref.
Although node->ref does not have to be accurate,
take this chance to use a more common READ_ONCE() and WRITE_ONCE()
pattern instead of data_race().
There is an existing bpf_lru_node_is_ref() and bpf_lru_node_set_ref().
This patch also adds bpf_lru_node_clear_ref() to do the
WRITE_ONCE(node->ref, 0) also.
==================================================================
BUG: KCSAN: data-race in __bpf_lru_list_rotate / __htab_lru_percpu_map_update_elem
read to 0xffff888137038deb of 1 bytes by task 11241 on cpu 0:
bpf_lru_node_set_ref kernel/bpf/bpf_lru_list.h:70 [inline]
__htab_lru_percpu_map_update_elem+0x2f1/0x820 kernel/bpf/hashtab.c:1332
bpf_percpu_hash_update+0x5e/0x90 kernel/bpf/hashtab.c:2313
bpf_map_update_value+0x2a9/0x370 kernel/bpf/syscall.c:200
generic_map_update_batch+0x3ae/0x4f0 kernel/bpf/syscall.c:1687
bpf_map_do_batch+0x2d9/0x3d0 kernel/bpf/syscall.c:4534
__sys_bpf+0x338/0x810
__do_sys_bpf kernel/bpf/syscall.c:5096 [inline]
__se_sys_bpf kernel/bpf/syscall.c:5094 [inline]
__x64_sys_bpf+0x43/0x50 kernel/bpf/syscall.c:5094
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
value changed: 0x01 -> 0x00
Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 11241 Comm: syz-executor.3 Not tainted 6.3.0-rc7-syzkaller-00136-g6a66fdd29ea1 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 03/30/2023
==================================================================
Reported-by: syzbot+ebe648a84e8784763f82@syzkaller.appspotmail.com Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/r/20230511043748.1384166-1-martin.lau@linux.dev Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Further investigation shows that the warning is superfluous, the migration
disabled task is just going to be migrated to its current running CPU.
This is because that on load balance if the dst_cpu is not allowed by the
task, we'll re-select a new_dst_cpu as a candidate. If no task can be
balanced to dst_cpu we'll try to balance the task to the new_dst_cpu
instead. In this case when the migration disabled task is not on CPU it
only allows to run on its current CPU, load balance will select its
current CPU as new_dst_cpu and later triggers the warning above.
The new_dst_cpu is chosen from the env->dst_grpmask. Currently it
contains CPUs in sched_group_span() and if we have overlapped groups it's
possible to run into this case. This patch makes env->dst_grpmask of
group_balance_mask() which exclude any CPUs from the busiest group and
solve the issue. For balancing in a domain with no overlapped groups
the behaviour keeps same as before.
Suggested-by: Vincent Guittot <vincent.guittot@linaro.org> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org> Link: https://lore.kernel.org/r/20230530082507.10444-1-yangyicong@huawei.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Both create_mapping_noalloc() and update_mapping_prot() sanity-check
their 'virt' parameter, but the check itself doesn't make much sense.
The condition used today appears to be a historical accident.
The sanity-check condition:
if ((virt >= PAGE_END) && (virt < VMALLOC_START)) {
[ ... warning here ... ]
return;
}
... can only be true for the KASAN shadow region or the module region,
and there's no reason to exclude these specifically for creating and
updateing mappings.
When arm64 support was first upstreamed in commit:
if ((virt >= VA_START) && (virt < VMALLOC_START)) {
[ ... warning here ... ]
return;
}
This appear to have been a thinko. The commit moved the linear map to
the bottom of the kernel address space, with VMALLOC_START being at the
halfway point. The old condition would warn for changes to the linear
map below this, and at the time VA_START was the end of the linear map.
Subsequently we cleaned up the naming of VA_START in commit:
if ((virt >= PAGE_END) && (virt < VMALLOC_START)) {
[ ... warning here ... ]
return;
}
Correct the condition to check against the start of the TTBR1 address
space, which is currently PAGE_OFFSET. This simplifies the logic, and
more clearly matches the "outside kernel range" message in the warning.
Signed-off-by: Mark Rutland <mark.rutland@arm.com> Cc: Russell King <linux@armlinux.org.uk> Cc: Steve Capper <steve.capper@arm.com> Cc: Will Deacon <will@kernel.org> Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Link: https://lore.kernel.org/r/20230615102628.1052103-1-mark.rutland@arm.com Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
posix_timer_add() tries to allocate a posix timer ID by starting from the
cached ID which was stored by the last successful allocation.
This is done in a loop searching the ID space for a free slot one by
one. The loop has to terminate when the search wrapped around to the
starting point.
But that's racy vs. establishing the starting point. That is read out
lockless, which leads to the following problem:
So CPU1 can observe a negative start value, i.e. -1, and the loop break
never happens because the condition can never be true:
if (sig->posix_timer_id == start)
break;
While this is unlikely to ever turn into an endless loop as the ID space is
huge (INT_MAX), the racy read of the start value caught the attention of
KCSAN and Dmitry unearthed that incorrectness.
Rewrite it so that all id operations are under the hash lock.
Currently, there is no limit for raid1/raid10 plugged bio. While flushing
writes, raid1 has cond_resched() while raid10 doesn't, and too many
writes can cause soft lockup.
Follow up soft lockup can be triggered easily with writeback test for
raid10 with ramdisks:
Fix the problem by adding cond_resched() to raid10 like what raid1 did.
Note that unlimited plugged bio still need to be optimized, for example,
in the case of lots of dirty pages writeback, this will take lots of
memory and io will spend a long time in plug, hence io latency is bad.
If the index allocated by idr_alloc greater than MINORMASK >> part_shift,
the device number will overflow, resulting in failure to create a block
device.
When run on a file system where the inline_data feature has been
enabled, xfstests generic/269, generic/270, and generic/476 cause ext4
to emit error messages indicating that inline directory entries are
corrupted. This occurs because the inline offset used to locate
inline directory entries in the inode body is not updated when an
xattr in that shared region is deleted and the region is shifted in
memory to recover the space it occupied. If the deleted xattr precedes
the system.data attribute, which points to the inline directory entries,
that attribute will be moved further up in the region. The inline
offset continues to point to whatever is located in system.data's former
location, with unfortunate effects when used to access directory entries
or (presumably) inline data in the inode body.
dmt_mode is allocated and never freed in this function.
It was found with the ast driver, but most drivers using generic fbdev
setup are probably affected.
When looking for something else in LKFT reports [1], I noticed that the
TC selftest ended with a timeout error:
not ok 1 selftests: tc-testing: tdc.sh # TIMEOUT 45 seconds
The timeout had been introduced 3 years ago, see the Fixes commit below.
This timeout is only in place when executing the selftests via the
kselftests runner scripts. I guess this is not what most TC devs are
using and nobody noticed the issue before.
The new timeout is set to 15 minutes as suggested by Pedro [2]. It looks
like it is plenty more time than what it takes in "normal" conditions.
If the LOOKUP request triggered from fuse_dentry_revalidate() is
interrupted, then the dentry will be invalidated, possibly resulting in
submounts being unmounted.
The warning happens because btrfs_qgroup_destroy_extent_records(), called
in the transaction abort path, we free all entries from the rbtree
"dirty_extent_root" with rbtree_postorder_for_each_entry_safe(), but we
don't actually empty the rbtree - it's still pointing to nodes that were
freed.
So set the rbtree's root node to NULL to avoid this warning (assign
RB_ROOT).
Fixes: 81f7eb00ff5b ("btrfs: destroy qgroup extent records on transaction abort") CC: stable@vger.kernel.org # 5.10+ Reviewed-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The trouble is that the drm_dev_unplugged() checks are by design racy,
they do not synchronize against all outstanding ioctl. This is because
those ioctl could block forever (both for modeset and for driver
specific ioctls), leading to deadlocks in hotunplug. Instead the code
sections that touch the hardware need to be annotated with
drm_dev_enter/exit, to avoid accessing hardware resources after the
unload/remove has finished.
To avoid use-after-free issues all the involved userspace visible
objects are supposed to hold a reference on the underlying drm_device,
like drm_file does.
The issue now is that we missed one, the atomic modeset ioctl can be run
in a nonblocking fashion, and in that case it cannot rely on the implied
drm_device reference provided by the ioctl calling context. This can
result in a use-after-free if an nonblocking atomic commit is carefully
raced against a driver unload.
Fix this by unconditionally grabbing a drm_device reference for any
drm_atomic_state structures. Strictly speaking this isn't required for
blocking commits and TEST_ONLY calls, but it's the simpler approach.
Thanks to shanzhulig for the initial idea of grabbing an unconditional
reference, I just added comments, a condensed commit message and fixed a
minor potential issue in where exactly we drop the final reference.
Reported-by: shanzhulig <shanzhulig@gmail.com> Suggested-by: shanzhulig <shanzhulig@gmail.com> Reviewed-by: Maxime Ripard <mripard@kernel.org> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: David Airlie <airlied@gmail.com> Cc: stable@kernel.org Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
System crash due to use after free.
Current code allows terminate_rport_io to exit before making
sure all IOs has returned. For FCP-2 device, IO's can hang
on in HW because driver has not tear down the session in FW at
first sign of cable pull. When dev_loss_tmo timer pops,
terminate_rport_io is called and upper layer is about to
free various resources. Terminate_rport_io trigger qla to do
the final cleanup, but the cleanup might not be fast enough where it
leave qla still holding on to the same resource.
Wait for IO's to return to upper layer before resources are freed.
Fix not to count the error code (which is minus value) to the total
used length of array, because it can mess up the return code of
process_fetch_insn_bottom(). Also clear the 'ret' value because it
will be used for calculating next data_loc entry.
Fix an issue in function 'tracing_err_log_open'.
The function doesn't call 'seq_open' if the file is opened only with
write permissions, which results in 'file->private_data' being left as null.
If we then use 'lseek' on that opened file, 'seq_lseek' dereferences
'file->private_data' in 'mutex_lock(&m->lock)', resulting in a kernel panic.
Writing to this node requires root privileges, therefore this bug
has very little security impact.
Tracefs node: /sys/kernel/tracing/error_log
Example Kernel panic:
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000038
Call trace:
mutex_lock+0x30/0x110
seq_lseek+0x34/0xb8
__arm64_sys_lseek+0x6c/0xb8
invoke_syscall+0x58/0x13c
el0_svc_common+0xc4/0x10c
do_el0_svc+0x24/0x98
el0_svc+0x24/0x88
el0t_64_sync_handler+0x84/0xe4
el0t_64_sync+0x1b4/0x1b8
Code: d503201faa0803e0aa1f03e1aa0103e9 (c8e97d02)
---[ end trace 561d1b49c12cf8a5 ]---
Kernel panic - not syncing: Oops: Fatal exception
split_if_spec expects a NULL-pointer as an end marker for the argument
list, but tuntap_probe never supplied that terminating NULL. As a result
incorrectly formatted interface specification string may cause a crash
because of the random memory access. Fix that by adding NULL terminator
to the split_if_spec argument list.
Cc: stable@vger.kernel.org Fixes: 7282bee78798 ("[PATCH] xtensa: Architecture support for Tensilica Xtensa Part 8") Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Through the vmcore, I found it's because in tracing_read_pipe(),
ring_buffer_empty_cpu() found some buffer is not empty but then it
cannot read anything due to "rb_num_of_entries() == 0" always true,
Then it infinitely loop the procedure due to user buffer not been
filled, see following code path:
tracing_read_pipe() {
... ...
waitagain:
tracing_wait_pipe() // 1. find non-empty buffer here
trace_find_next_entry_inc() // 2. loop here try to find an entry
__find_next_entry()
ring_buffer_empty_cpu(); // 3. find non-empty buffer
peek_next_entry() // 4. but peek always return NULL
ring_buffer_peek()
rb_buffer_peek()
rb_get_reader_page()
// 5. because rb_num_of_entries() == 0 always true here
// then return NULL
// 6. user buffer not been filled so goto 'waitgain'
// and eventually leads to an deadloop in kernel!!!
}
By some analyzing, I found that when resetting ringbuffer, the 'entries'
of its pages are not all cleared (see rb_reset_cpu()). Then when reducing
the ringbuffer, and if some reduced pages exist dirty 'entries' data, they
will be added into 'cpu_buffer->overrun' (see rb_remove_pages()), which
cause wrong 'overrun' count and eventually cause the deadloop issue.
To fix it, we need to clear every pages in rb_reset_cpu().
Hist triggers can have referenced variables without having direct
variables fields. This can be the case if referenced variables are added
for trigger actions. In this case the newly added references will not
have field variables. Not taking such referenced variables into
consideration can result in a bug where it would be possible to remove
hist trigger with variables being refenced. This will result in a bug
that is easily reproducable like so
We hit the bug because when second hist trigger has was created
has_hist_vars() returned false because hist trigger did not have
variables. As a result of that save_hist_vars() was not called to add
the trigger to trace_array->hist_vars. Later on when we attempted to
remove the first histogram find_any_var_ref() failed to detect it is
being used because it did not find the second trigger in hist_vars list.
With this change we wait until trigger actions are created so we can take
into consideration if hist trigger has variable references. Also, now we
check the return value of save_hist_vars() and fail trigger creation if
save_hist_vars() fails.
When the best clk is searched, we iterate over all possible clk.
If we find a better match, the previous one, if any, needs to be freed.
If a better match has already been found, we still need to free the new
one, otherwise it leaks.
Commit eb26dfe8aa7e ("8250: add support for ASIX devices with a FIFO
bug") merged on Jul 13, 2012 adds a quirk for PCI_VENDOR_ID_ASIX
(0x9710). But that ID is the same as PCI_VENDOR_ID_NETMOS defined in 1f8b061050c7 ("[PATCH] Netmos parallel/serial/combo support") merged
on Mar 28, 2005. In pci_serial_quirks array, the NetMos entry always
takes precedence over the ASIX entry even since it was initially
merged, code in that commit is always unreachable.
In my tests, adding the FIFO workaround to pci_netmos_init() makes no
difference, and the vendor driver also does not have such workaround.
Given that the code was never used for over a decade, it's safe to
revert it.
Also, the real PCI_VENDOR_ID_ASIX should be 0x125b, which is used on
their newer AX99100 PCIe serial controllers released on 2016. The FIFO
workaround should not be intended for these newer controllers, and it
was never implemented in vendor driver.
Fixes: eb26dfe8aa7e ("8250: add support for ASIX devices with a FIFO bug") Cc: stable <stable@kernel.org> Signed-off-by: Jiaqing Zhao <jiaqing.zhao@linux.intel.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20230619155743.827859-1-jiaqing.zhao@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to the datasheets of supported meson SoCs length of ADC_CLK_DIV
field is 6-bit. Although all supported SoCs have the register
with that field documented later SoCs use external clock rather than
ADC internal clock so this patch affects only meson8 family (S8* SoCs).
Fixes: 3adbf3427330 ("iio: adc: add a driver for the SAR ADC found in Amlogic Meson SoCs") Signed-off-by: George Stark <GNStark@sberdevices.ru> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Link: https://lore.kernel.org/r/20230606165357.42417-1-gnstark@sberdevices.ru Cc: <stable@vger.kernel.org> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a client sends out a cap update dropping caps with the prior 'seq'
just before an incoming cap revoke request, then the client may drop
the revoke because it believes it's already released the requested
capabilities.
This causes the MDS to wait indefinitely for the client to respond
to the revoke. It's therefore always a good idea to ack the cap
revoke request with the bumped up 'seq'.
Fix the timeout that is used for the initialisation and for the self
test. wait_for_completion_timeout expects a timeout in jiffies, but
RNGC_TIMEOUT is in milliseconds. Call msecs_to_jiffies to do the
conversion.
Cc: stable@vger.kernel.org Fixes: 1d5449445bd0 ("hwrng: mx-rngc - add a driver for Freescale RNGC") Signed-off-by: Martin Kaiser <martin@kaiser.cx> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The atmel_complete_tx_dma() function disables IRQs at the start
of the function by calling spin_lock_irqsave(&port->lock, flags);
There is no need to disable them a second time using the
spin_lock_irq() function and, in fact, doing so is a bug because
it will enable IRQs prematurely when we call spin_unlock_irq().
Just use spin_lock/unlock() instead without disabling or enabling
IRQs.
If we disable vblank when entering self-refresh, vblank APIs (like
DRM_IOCTL_WAIT_VBLANK) no longer work. But user space is not aware when
we enter self-refresh, so this appears to be an API violation -- that
DRM_IOCTL_WAIT_VBLANK fails with EINVAL whenever the display is idle and
enters self-refresh.
The downstream driver used by many of these systems never used to
disable vblank for PSR, and in fact, even upstream, we didn't do that
until radically redesigning the state machine in commit 6c836d965bad
("drm/rockchip: Use the helpers for PSR").
Thus, it seems like a reasonable API fix to simply restore that
behavior, and leave vblank enabled.
Note that this appears to potentially unbalance the
drm_crtc_vblank_{off,on}() calls in some cases, but:
(a) drm_crtc_vblank_on() documents this as OK and
(b) if I do the naive balancing, I find state machine issues such that
we're not in sync properly; so it's easier to take advantage of (a).
This issue was exposed by IGT's kms_vblank tests, and reported by
KernelCI. The bug has been around a while (longer than KernelCI
noticed), but was only exposed once self-refresh was bugfixed more
recently, and so KernelCI could properly test it. Some other notes in:
https://lore.kernel.org/dri-devel/Y6OCg9BPnJvimQLT@google.com/
Re: renesas/master bisection: igt-kms-rockchip.kms_vblank.pipe-A-wait-forked on rk3399-gru-kevin
== Backporting notes: ==
Marking as 'Fixes' commit 6c836d965bad ("drm/rockchip: Use the helpers
for PSR"), but it probably depends on commit bed030a49f3e
("drm/rockchip: Don't fully disable vop on self refresh") as well.
We also need the previous patch ("drm/atomic: Allow vblank-enabled +
self-refresh "disable""), of course.
The self-refresh helper framework overloads "disable" to sometimes mean
"go into self-refresh mode," and this mode activates automatically
(e.g., after some period of unchanging display output). In such cases,
the display pipe is still considered "on", and user-space is not aware
that we went into self-refresh mode. Thus, users may expect that
vblank-related features (such as DRM_IOCTL_WAIT_VBLANK) still work
properly.
However, we trigger the WARN_ONCE() here if a CRTC driver tries to leave
vblank enabled.
Add a different expectation: that CRTCs *should* leave vblank enabled
when going into self-refresh.
This patch is preparation for another patch -- "drm/rockchip: vop: Leave
vblank enabled in self-refresh" -- which resolves conflicts between the
above self-refresh behavior and the API tests in IGT's kms_vblank test
module.
== Some alternatives discussed: ==
It's likely that on many display controllers, vblank interrupts will
turn off when the CRTC is disabled, and so in some cases, self-refresh
may not support vblank. To support such cases, we might consider
additions to the generic helpers such that we fire vblank events based
on a timer.
However, there is currently only one driver using the common
self-refresh helpers (i.e., rockchip), and at least as of commit bed030a49f3e ("drm/rockchip: Don't fully disable vop on self refresh"),
the CRTC hardware is powered enough to continue to generate vblank
interrupts.
So we chose the simpler option of leaving vblank interrupts enabled. We
can reevaluate this decision and perhaps augment the helpers if/when we
gain a second driver that has different requirements.
v3:
* include discussion summary
v2:
* add 'ret != 0' warning case for self-refresh
* describe failing test case and relation to drm/rockchip patch better
The GETLK pid values have all been negated since commit 9d5b86ac13c5
("fs/locks: Remove fl_nspid and use fs-specific l_pid for remote locks").
Revert this for local pids, and leave in place negative pids for remote
owners.
Cc: stable@vger.kernel.org Fixes: 9d5b86ac13c5 ("fs/locks: Remove fl_nspid and use fs-specific l_pid for remote locks") Signed-off-by: Alexander Aring <aahringo@redhat.com> Signed-off-by: David Teigland <teigland@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We've found that using raid0 with the 'original' layout and discard
enabled with different disk sizes (such that at least two zones are
created) can result in data corruption. This is due to the fact that
the discard handling in 'raid0_handle_discard()' assumes the 'alternate'
layout. We've seen this corruption using ext4 but other filesystems are
likely susceptible as well.
More specifically, while multiple zones are necessary to create the
corruption, the corruption may not occur with multiple zones if they
layout in such a way the layout matches what the 'alternate' layout
would have produced. Thus, not all raid0 devices with the 'original'
layout, different size disks and discard enabled will encounter this
corruption.
The 3.14 kernel inadvertently changed the raid0 disk layout for different
size disks. Thus, running a pre-3.14 kernel and post-3.14 kernel on the
same raid0 array could corrupt data. This lead to the creation of the
'original' layout (to match the pre-3.14 layout) and the 'alternate' layout
(to match the post 3.14 layout) in the 5.4 kernel time frame and an option
to tell the kernel which layout to use (since it couldn't be autodetected).
However, when the 'original' layout was added back to 5.4 discard support
for the 'original' layout was not added leading this issue.
I've been able to reliably reproduce the corruption with the following
test case:
1. create raid0 array with different size disks using original layout
2. mkfs
3. mount -o discard
4. create lots of files
5. remove 1/2 the files
6. fstrim -a (or just the mount point for the raid0 array)
7. umount
8. fsck -fn /dev/md0 (spews all sorts of corruptions)
Let's fix this by adding proper discard support to the 'original' layout.
The fix 'maps' the 'original' layout disks to the order in which they are
read/written such that we can compare the disks in the same way that the
current 'alternate' layout does. A 'disk_shift' field is added to
'struct strip_zone'. This could be computed on the fly in
raid0_handle_discard() but by adding this field, we save some computation
in the discard path.
Note we could also potentially fix this by re-ordering the disks in the
zones that follow the first one, and then always read/writing them using
the 'alternate' layout. However, that is seen as a more substantial change,
and we are attempting the least invasive fix at this time to remedy the
corruption.
I've verified the change using the reproducer mentioned above. Typically,
the corruption is seen after less than 3 iterations, while the patch has
run 500+ iterations.
Cc: NeilBrown <neilb@suse.de> Cc: Song Liu <song@kernel.org> Fixes: c84a1372df92 ("md/raid0: avoid RAID0 data corruption due to layout confusion.") Cc: stable@vger.kernel.org Signed-off-by: Jason Baron <jbaron@akamai.com> Signed-off-by: Song Liu <song@kernel.org> Link: https://lore.kernel.org/r/20230623180523.1901230-1-jbaron@akamai.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The irq_raised completion used to detect the end of a test case is
initialized when the test device is probed, but never reinitialized again
before a test case. As a result, the irq_raised completion synchronization
is effective only for the first ioctl test case executed. Any subsequent
call to wait_for_completion() by another ioctl() call will immediately
return, potentially too early, leading to false positive failures.
Fix this by reinitializing the irq_raised completion before starting a new
ioctl() test command.
Link: https://lore.kernel.org/r/20230415023542.77601-16-dlemoal@kernel.org Fixes: 2c156ac71c6b ("misc: Add host side PCI driver for PCI test function device") Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Manivannan Sadhasivam <mani@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In pci_endpoint_test_remove(), freeing the IRQs after removing the device
creates a small race window for IRQs to be received with the test device
memory already released, causing the IRQ handler to access invalid memory,
resulting in an oops.
Free the device IRQs before removing the device to avoid this issue.
The address translation unit of the rockchip EP controller does not use
the lower 8 bits of a PCIe-space address to map local memory. Thus we
must set the align feature field to 256 to let the user know about this
constraint.
Link: https://lore.kernel.org/r/20230418074700.1083505-12-rick.wertenbroek@gmail.com Fixes: cf590b078391 ("PCI: rockchip: Add EP driver for Rockchip PCIe controller") Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Rick Wertenbroek <rick.wertenbroek@gmail.com> Signed-off-by: Lorenzo Pieralisi <lpieralisi@kernel.org> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>