The device_for_each_child_node() macro requires explicit calls to
fwnode_handle_put() in all early exits of the loop if the child node is
not required outside. Otherwise, the child node's refcount is not
decremented and the resource is not released.
The current implementation of pmic_glink_ucsi_probe() makes use of the
device_for_each_child_node(), but does not release the child node on
early returns. Add the missing calls to fwnode_handle_put().
Cc: stable@vger.kernel.org Fixes: c6165ed2f425 ("usb: ucsi: glink: use the connector orientation GPIO to provide switch events") Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Link: https://lore.kernel.org/r/20240613-ucsi-glink-release-node-v1-1-f7629a56f70a@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the aspeed UDC setup, we configure the UDC hardware with the assigned
USB device address.
However, we have an off-by-one in the bitmask, so we're only setting the
lower 6 bits of the address (USB addresses being 7 bits, and the
hardware bitmask being bits 0:6).
This means that device enumeration fails if the assigned address is
greater than 64:
[ 344.607255] usb 1-1: new high-speed USB device number 63 using ehci-platform
[ 344.808459] usb 1-1: New USB device found, idVendor=cc00, idProduct=cc00, bcdDevice= 6.10
[ 344.817684] usb 1-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3
[ 344.825671] usb 1-1: Product: Test device
[ 344.831075] usb 1-1: Manufacturer: Test vendor
[ 344.836335] usb 1-1: SerialNumber: 00
[ 349.917181] usb 1-1: USB disconnect, device number 63
[ 352.036775] usb 1-1: new high-speed USB device number 64 using ehci-platform
[ 352.249432] usb 1-1: device descriptor read/all, error -71
[ 352.696740] usb 1-1: new high-speed USB device number 65 using ehci-platform
[ 352.909431] usb 1-1: device descriptor read/all, error -71
Use the correct mask of 0x7f (rather than 0x3f), and generate this
through the GENMASK macro, so we have numbers that correspond exactly
to the hardware register definition.
When config CONFIG_USB_DWC3_DUAL_ROLE is selected, and trigger system
to enter suspend status with below command:
echo mem > /sys/power/state
There will be a deadlock issue occurring. Detailed invoking path as
below:
dwc3_suspend_common()
spin_lock_irqsave(&dwc->lock, flags); <-- 1st
dwc3_gadget_suspend(dwc);
dwc3_gadget_soft_disconnect(dwc);
spin_lock_irqsave(&dwc->lock, flags); <-- 2nd
This issue is exposed by commit c7ebd8149ee5 ("usb: dwc3: gadget: Fix
NULL pointer dereference in dwc3_gadget_suspend") that removes the code
of checking whether dwc->gadget_driver is NULL or not. It causes the
following code is executed and deadlock occurs when trying to get the
spinlock. In fact, the root cause is the commit 5265397f9442("usb: dwc3:
Remove DWC3 locking during gadget suspend/resume") that forgot to remove
the lock of otg mode. So, remove the redundant lock of otg mode during
gadget suspend/resume.
Fixes: 5265397f9442 ("usb: dwc3: Remove DWC3 locking during gadget suspend/resume") Cc: Xu Yang <xu.yang_2@nxp.com> Cc: stable@vger.kernel.org Signed-off-by: Meng Li <Meng.Li@windriver.com> Acked-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com> Link: https://lore.kernel.org/r/20240618031918.2585799-1-Meng.Li@windriver.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Syzbot is still reporting quite an old issue [1] that occurs due to
incomplete checking of present usb endpoints. As such, wrong
endpoints types may be used at urb sumbitting stage which in turn
triggers a warning in usb_submit_urb().
Fix the issue by verifying that required endpoint types are present
for both in and out endpoints, taking into account cmd endpoint type.
Unfortunately, this patch has not been tested on real hardware.
printer_read() and printer_write() guard against the race
against disable() by checking the dev->interface flag,
which in turn is guarded by a spinlock.
These functions, however, drop the lock on multiple occasions.
This means that the test has to be redone after reacquiring
the lock and before doing IO.
Add the tests.
This also addresses CVE-2024-25741
Fixes: 7f2ca14d2f9b9 ("usb: gadget: function: printer: Interface is disabled and returns error") Cc: stable <stable@kernel.org> Signed-off-by: Oliver Neukum <oneukum@suse.com> Link: https://lore.kernel.org/r/20240620114039.5767-1-oneukum@suse.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Avoid spurious link status logs that may ultimately be wrong; for example,
if the link is set to down with the cable plugged, then the cable is
unplugged and after this the link is set to up, the last new log that is
appearing is incorrectly telling that the link is up.
In order to avoid errors, show link status logs after link_reset
processing, and in order to avoid spurious as much as possible, only show
the link loss when some link status change is detected.
cc: stable@vger.kernel.org Fixes: e2ca90c276e1 ("ax88179_178a: ASIX AX88179_178A USB 3.0/2.0 to gigabit ethernet adapter driver") Signed-off-by: Jose Ignacio Tornos Martinez <jtornosm@redhat.com> Reviewed-by: Simon Horman <horms@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
a) Set sensor to forced mode
b) Sensor measures values and update data registers and sleeps again
c) Read data registers
In the current implementation the read operation happens immediately
after the sensor is set to forced mode so the sensor does not have
the time to update properly the registers. This leads to the following
2 problems:
1) The first ever value which is read by the register is always wrong
2) Every read operation, puts the register into forced mode and reads
the data that were calculated in the previous conversion.
This behaviour was tested in 2 ways:
1) The internal meas_status_0 register was read before and after every
read operation in order to verify that the data were ready even before
the register was set to forced mode and also to check that after the
forced mode was set the new data were not yet ready.
2) Physically changing the temperature and measuring the temperature
This commit adds the waiting time in between the set of the forced mode
and the read of the data. The function is taken from the Bosch BME68x
Sensor API [1].
There are cases in the compensate functions of the driver that
there could be overflows of variables due to bit shifting ops.
These implications were initially discussed here [1] and they
were mentioned in log message of Commit 1b3bd8592780 ("iio:
chemical: Add support for Bosch BME680 sensor").
According to the ABI docs hysteresis values are represented as offsets to
threshold values. Current implementation represents hysteresis values as
absolute values which is wrong. Nevertheless the device stores them as
absolute values and the datasheet refers to them as clear thresholds. Fix
the reading and writing of hysteresis values by including thresholds into
calculations. Hysteresis values that result in threshold clear values
that are out of limits will be truncated.
To check that the threshold clear values are correct, registers are read
out using i2ctransfer and the corresponding temperature and relative
humidity thresholds are calculated using the formulas in the datasheet.
libsas is currently not freeing all the struct ata_port struct members,
e.g. ncq_sense_buf for a driver supporting Command Duration Limits (CDL).
Add a function, ata_port_free(), that is used to free a ata_port,
including its struct members. It makes sense to keep the code related to
freeing a ata_port in its own function, which will also free all the
struct members of struct ata_port.
Fixes: 18bd7718b5c4 ("scsi: ata: libata: Handle completion of CDL commands using policy 0xD") Reviewed-by: John Garry <john.g.garry@oracle.com> Link: https://lore.kernel.org/r/20240629124210.181537-8-cassel@kernel.org Signed-off-by: Niklas Cassel <cassel@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When clearing registers on new write requests was added, the protection
for currently running commands was missed leading to concurrent access
to the testunit registers. Check the flag beforehand.
Fixes: b39ab96aa894 ("i2c: testunit: add support for block process calls") Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Andi Shyti <andi.shyti@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
STOP fallsthrough to WRITE_REQUESTED but this became problematic when
clearing the testunit registers was added to the latter. Actually, there
is no reason to clear the testunit state after STOP. Doing it when a new
WRITE_REQUESTED arrives is enough. So, no need to fallthrough, at all.
Fixes: b39ab96aa894 ("i2c: testunit: add support for block process calls") Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Andi Shyti <andi.shyti@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The make deb-pkg target calls debian-orig which attempts to either
hard link the source .tar to the build-output location or copy the
source .tar to the build-output location. The test to determine
whether to ln or cp is incorrectly expanded by Make and consequently
always attempts to ln the source .tar. This fix corrects the escaping
of '$' so that the test is expanded by the shell rather than by Make
and appropriately selects between ln and cp.
Fixes: b44aa8c96e9e ("kbuild: deb-pkg: make .orig tarball a hard link if possible") Signed-off-by: Thayne Harbaugh <thayne@mastodonlabs.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The default INSTALL_MOD_DIR was changed from 'extra' to
'updates' in commit b74d7bb7ca24 ("kbuild: Modify default
INSTALL_MOD_DIR from extra to updates").
This commit updates the documentation to align with the
latest kernel.
Fixes: b74d7bb7ca24 ("kbuild: Modify default INSTALL_MOD_DIR from extra to updates") Signed-off-by: Mark-PK Tsai <mark-pk.tsai@mediatek.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
I still see "RPC: Could not send backchannel reply error: -110"
quite often, along with slow-running tests. Debugging shows that the
backchannel is still stumbling when it has to queue a callback reply
on a busy transport.
Note that every one of these timeouts causes a connection loss by
virtue of the xprt_conditional_disconnect() call in that arm of
call_cb_transmit_status().
I found that setting to_maxval is necessary to get the RPC timeout
logic to behave whenever to_exponential is not set.
Fixes: 57331a59ac0d ("NFSv4.1: Use the nfs_client's rpc timeouts for backchannel") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
ams_enable_channel_sequence constructs a "scan_mask" for all the PS and
PL channels. This works out fine, since scan_index for these channels is
less than 64. However, it also includes the ams_ctrl_channels, where
scan_index is greater than 64, triggering undefined behavior. Since we
don't need these channels anyway, just exclude them.
Fixes: d5c70627a794 ("iio: adc: Add Xilinx AMS driver") Signed-off-by: Sean Anderson <sean.anderson@linux.dev> Link: https://lore.kernel.org/r/20240311162800.11074-1-sean.anderson@linux.dev Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
sdhci_check_ro() can call mmc_gpio_get_ro() while holding the sdhci
host->lock spinlock. That would be a problem if the GPIO access done by
mmc_gpio_get_ro() needed to sleep.
However, host->lock is not needed anyway. The mmc core ensures that host
operations do not race with each other, and asynchronous callbacks like the
interrupt handler, software timeouts, completion work etc, cannot affect
sdhci_check_ro().
So remove the locking.
Fixes: 6d5cd068ee59 ("mmc: sdhci: use WP GPIO in sdhci_check_ro()") Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20240614080051.4005-3-adrian.hunter@intel.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mmc_of_parse() reads device property "wp-inverted" and sets
MMC_CAP2_RO_ACTIVE_HIGH if it is true. MMC_CAP2_RO_ACTIVE_HIGH is used
to invert a write-protect (AKA read-only) GPIO value.
sdhci_get_property() also reads "wp-inverted" and sets
SDHCI_QUIRK_INVERTED_WRITE_PROTECT which is used to invert the
write-protect value as well but also acts upon a value read out from the
SDHCI_PRESENT_STATE register.
Many drivers call both mmc_of_parse() and sdhci_get_property(),
so that both MMC_CAP2_RO_ACTIVE_HIGH and
SDHCI_QUIRK_INVERTED_WRITE_PROTECT will be set if the controller has
device property "wp-inverted".
Amend the logic in sdhci_check_ro() to allow for that possibility,
so that the write-protect value is not inverted twice.
Also do not invert the value if it is a negative error value. Note that
callers treat an error the same as not-write-protected, so the result is
functionally the same in that case.
Also do not invert the value if sdhci host operation ->get_ro() is used.
None of the users of that callback set SDHCI_QUIRK_INVERTED_WRITE_PROTECT
directly or indirectly, but two do call mmc_gpio_get_ro(), so leave it to
them to deal with that if they ever set SDHCI_QUIRK_INVERTED_WRITE_PROTECT
in the future.
Fixes: 6d5cd068ee59 ("mmc: sdhci: use WP GPIO in sdhci_check_ro()") Signed-off-by: Adrian Hunter <adrian.hunter@intel.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20240614080051.4005-2-adrian.hunter@intel.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
jmicron_pmos() and sdhci_pci_probe() use pci_{read,write}_config_byte()
that return PCIBIOS_* codes. The return code is then returned as is by
jmicron_probe() and sdhci_pci_probe(). Similarly, the return code is
also returned as is from jmicron_resume(). Both probe and resume
functions should return normal errnos.
Convert PCIBIOS_* returns code using pcibios_err_to_errno() into normal
errno before returning them the fix these issues.
When erase/trim/discard completion was converted to mmc_poll_for_busy(),
optional support to poll with the host_ops->card_busy() callback was also
added.
The common sdhci's ->card_busy() turns out not to be working as expected
for the sdhci-brcmstb variant, as it keeps returning busy beyond the card's
busy period. In particular, this leads to the below splat for
mmc_do_erase() when running a discard (BLKSECDISCARD) operation during
mkfs.f2fs:
To fix the problem let's override the host_ops->card_busy() callback by
setting it to NULL, which forces the mmc core to poll with a CMD13 and
checking the R1_STATUS in the mmc_busy_cb() function.
Signed-off-by: Kamal Dasu <kamal.dasu@broadcom.com> Fixes: 0d84c3e6a5b2 ("mmc: core: Convert to mmc_poll_for_busy() for erase/trim/discard") Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20240603220834.21989-2-kamal.dasu@broadcom.com
[Ulf: Clarified the commit message] Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
sdhci_pci_o2_probe() uses pci_read_config_{byte,dword}() that return
PCIBIOS_* codes. The return code is then returned as is but as
sdhci_pci_o2_probe() is probe function chain, it should return normal
errnos.
Convert PCIBIOS_* returns code using pcibios_err_to_errno() into normal
errno before returning them. Add a label for read failure so that the
conversion can be done in one place rather than on all of the return
statements.
Fixes: 3d757ddbd68c ("mmc: sdhci-pci-o2micro: add Bayhub new chip GG8 support for UHS-I") Fixes: d599005afde8 ("mmc: sdhci-pci-o2micro: Add missing checks in sdhci_pci_o2_probe") Fixes: 706adf6bc31c ("mmc: sdhci-pci-o2micro: Add SeaBird SeaEagle SD3 support") Fixes: 01acf6917aed ("mmc: sdhci-pci: add support of O2Micro/BayHubTech SD hosts") Fixes: 26daa1ed40c6 ("mmc: sdhci: Disable ADMA on some O2Micro SD/MMC parts.") Cc: stable@vger.kernel.org Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Link: https://lore.kernel.org/r/20240527132443.14038-2-ilpo.jarvinen@linux.intel.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The requirement that the head page be passed to do_set_pmd() was added in
commit ef37b2ea08ac ("mm/memory: page_add_file_rmap() ->
folio_add_file_rmap_[pte|pmd]()") and prevents pmd-mapping in the
finish_fault() and filemap_map_pages() paths if the page to be inserted is
anything but the head page for an otherwise suitable vma and pmd-sized
page.
Matthew said:
: We're going to stop using PMDs to map large folios unless the fault is
: within the first 4KiB of the PMD. No idea how many workloads that
: affects, but it only needs to be backported as far as v6.8, so we may
: as well backport it.
Link: https://lkml.kernel.org/r/20240611153216.2794513-1-abrestic@rivosinc.com Fixes: ef37b2ea08ac ("mm/memory: page_add_file_rmap() -> folio_add_file_rmap_[pte|pmd]()") Signed-off-by: Andrew Bresticker <abrestic@rivosinc.com> Acked-by: David Hildenbrand <david@redhat.com> Acked-by: Hugh Dickins <hughd@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
xa_for_each() in _vm_unmap_aliases() loops through all vbs. However,
since commit 062eacf57ad9 ("mm: vmalloc: remove a global vmap_blocks
xarray") the vb from xarray may not be on the corresponding CPU
vmap_block_queue. Consequently, purge_fragmented_block() might use the
wrong vbq->lock to protect the free list, leading to vbq->free breakage.
Incorrect lock protection can exhaust all vmalloc space as follows:
CPU0 CPU1
+--------------------------------------------+
| +--------------------+ +-----+ |
+--> | |---->| |------+
| CPU1:vbq free_list | | vb1 |
+--- | |<----| |<-----+
| +--------------------+ +-----+ |
+--------------------------------------------+
_vm_unmap_aliases() vb_alloc()
new_vmap_block()
xa_for_each(&vbq->vmap_blocks, idx, vb)
--> vb in CPU1:vbq->freelist
purge_fragmented_block(vb)
spin_lock(&vbq->lock) spin_lock(&vbq->lock)
--> use CPU0:vbq->lock --> use CPU1:vbq->lock
prev->next = next
+--------------------------------------------+
|----------------------------+ |
| +--------------------+ | +-----+ |
+--> | |--+ | |------+
| CPU1:vbq free_list | | vb2 |
+--- | |<----| |<-----+
| +--------------------+ +-----+ |
+--------------------------------------------+
Here’s a list breakdown. All vbs, which were to be added to
‘prev’, cannot be used by list_for_each_entry_rcu(vb, &vbq->free,
free_list) in vb_alloc(). Thus, vmalloc space is exhausted.
Commit 29d7355a9d05 ("kasan: save alloc stack traces for mempool") messed
up one of the calls to unpoison_slab_object: the last two arguments are
supposed to be GFP flags and whether to init the object memory.
Fix the call.
Without this fix, __kasan_mempool_unpoison_object provides the object's
size as GFP flags to unpoison_slab_object, which can cause LOCKDEP reports
(and probably other issues).
Link: https://lkml.kernel.org/r/20240614143238.60323-1-andrey.konovalov@linux.dev Fixes: 29d7355a9d05 ("kasan: save alloc stack traces for mempool") Signed-off-by: Andrey Konovalov <andreyknvl@gmail.com> Reported-by: Brad Spengler <spender@grsecurity.net> Acked-by: Marco Elver <elver@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since commit 2282679fb20b ("mm: submit multipage write for SWP_FS_OPS
swap-space"), we can plug multiple pages then unplug them all together.
That means iov_iter_count(iter) could be way bigger than PAGE_SIZE, it
actually equals the size of iov_iter_npages(iter, INT_MAX).
Note this issue has nothing to do with large folios as we don't support
THP_SWPOUT to non-block devices.
[v-songbaohua@oppo.com: figure out the cause and correct the commit message] Link: https://lkml.kernel.org/r/20240618065647.21791-1-21cnbao@gmail.com Fixes: 2282679fb20b ("mm: submit multipage write for SWP_FS_OPS swap-space") Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Barry Song <v-songbaohua@oppo.com> Closes: https://lore.kernel.org/linux-mm/20240617053201.GA16852@lst.de/ Reviewed-by: Martin Wege <martin.l.wege@gmail.com> Cc: NeilBrown <neilb@suse.de> Cc: Anna Schumaker <anna@kernel.org> Cc: Steve French <sfrench@samba.org> Cc: Trond Myklebust <trondmy@kernel.org> Cc: Chuanhua Han <hanchuanhua@oppo.com> Cc: Ryan Roberts <ryan.roberts@arm.com> Cc: Chris Li <chrisl@kernel.org> Cc: "Huang, Ying" <ying.huang@intel.com> Cc: Jeff Layton <jlayton@kernel.org> Cc: Matthew Wilcox <willy@infradead.org> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The code in ocfs2_dio_end_io_write() estimates number of necessary
transaction credits using ocfs2_calc_extend_credits(). This however does
not take into account that the IO could be arbitrarily large and can
contain arbitrary number of extents.
Extent tree manipulations do often extend the current transaction but not
in all of the cases. For example if we have only single block extents in
the tree, ocfs2_mark_extent_written() will end up calling
ocfs2_replace_extent_rec() all the time and we will never extend the
current transaction and eventually exhaust all the transaction credits if
the IO contains many single block extents. Once that happens a
WARN_ON(jbd2_handle_buffer_credits(handle) <= 0) is triggered in
jbd2_journal_dirty_metadata() and subsequently OCFS2 aborts in response to
this error. This was actually triggered by one of our customers on a
heavily fragmented OCFS2 filesystem.
To fix the issue make sure the transaction always has enough credits for
one extent insert before each call of ocfs2_mark_extent_written().
Heming Zhao said:
------
PANIC: "Kernel panic - not syncing: OCFS2: (device dm-1): panic forced after error"
Link: https://lkml.kernel.org/r/20240617095543.6971-1-jack@suse.cz Link: https://lkml.kernel.org/r/20240614145243.8837-1-jack@suse.cz Fixes: c15471f79506 ("ocfs2: fix sparse file & data ordering issue in direct io") Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> Reviewed-by: Heming Zhao <heming.zhao@suse.com> Cc: Mark Fasheh <mark@fasheh.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Junxiao Bi <junxiao.bi@oracle.com> Cc: Changwei Ge <gechangwei@live.cn> Cc: Gang He <ghe@suse.com> Cc: Jun Piao <piaojun@huawei.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The SPMI GPIO driver assumes that the parent device is an SPMI device
and accesses random data when backcasting the parent struct device
pointer for non-SPMI devices.
Fortunately this does not seem to cause any issues currently when the
parent device is an I2C client like the PM8008, but this could change if
the structures are reorganised (e.g. using structure randomisation).
Notably the interrupt implementation is also broken for non-SPMI devices.
Also note that the two GPIO pins on PM8008 are used for interrupts and
reset so their practical use should be limited.
Drop the broken GPIO support for PM8008 for now.
Fixes: ea119e5a482a ("pinctrl: qcom-pmic-gpio: Add support for pm8008") Cc: stable@vger.kernel.org # 5.13 Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org> Reviewed-by: Stephen Boyd <swboyd@chromium.org> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Link: https://lore.kernel.org/r/20240529162958.18081-9-johan+linaro@kernel.org Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The sys_fanotify_mark() syscall on parisc uses the reverse word order
for the two halves of the 64-bit argument compared to all syscalls on
all 32-bit architectures. As far as I can tell, the problem is that
the function arguments on parisc are sorted backwards (26, 25, 24, 23,
...) compared to everyone else, so the calling conventions of using an
even/odd register pair in native word order result in the lower word
coming first in function arguments, matching the expected behavior
on little-endian architectures. The system call conventions however
ended up matching what the other 32-bit architectures do.
A glibc cleanup in 2020 changed the userspace behavior in a way that
handles all architectures consistently, but this inadvertently broke
parisc32 by changing to the same method as everyone else.
The change made it into glibc-2.35 and subsequently into debian 12
(bookworm), which is the latest stable release. This means we
need to choose between reverting the glibc change or changing the
kernel to match it again, but either hange will leave some systems
broken.
Pick the option that is more likely to help current and future
users and change the kernel to match current glibc. This also
means the behavior is now consistent across architectures, but
it breaks running new kernels with old glibc builds before 2.35.
The 'profile_pc()' function is used for timer-based profiling, which
isn't really all that relevant any more to begin with, but it also ends
up making assumptions based on the stack layout that aren't necessarily
valid.
Basically, the code tries to account the time spent in spinlocks to the
caller rather than the spinlock, and while I support that as a concept,
it's not worth the code complexity or the KASAN warnings when no serious
profiling is done using timers anyway these days.
And the code really does depend on stack layout that is only true in the
simplest of cases. We've lost the comment at some point (I think when
the 32-bit and 64-bit code was unified), but it used to say:
Assume the lock function has either no stack frame or a copy
of eflags from PUSHF.
which explains why it just blindly loads a word or two straight off the
stack pointer and then takes a minimal look at the values to just check
if they might be eflags or the return pc:
Eflags always has bits 22 and up cleared unlike kernel addresses
but that basic stack layout assumption assumes that there isn't any lock
debugging etc going on that would complicate the code and cause a stack
frame.
It causes KASAN unhappiness reported for years by syzkaller [1] and
others [2].
With no real practical reason for this any more, just remove the code.
Just for historical interest, here's some background commits relating to
this code from 2006:
0cb91a229364 ("i386: Account spinlocks to the caller during profiling for !FP kernels") 31679f38d886 ("Simplify profile_pc on x86-64")
An unintended consequence of commit 9c573cd31343 ("randomize_kstack:
Improve entropy diffusion") was that the per-architecture entropy size
filtering reduced how many bits were being added to the mix, rather than
how many bits were being used during the offsetting. All architectures
fell back to the existing default of 0x3FF (10 bits), which will consume
at most 1KiB of stack space. It seems that this is working just fine,
so let's avoid the confusion and update everything to use the default.
The prior intent of the per-architecture limits were:
arm64: capped at 0x1FF (9 bits), 5 bits effective
powerpc: uncapped (10 bits), 6 or 7 bits effective
riscv: uncapped (10 bits), 6 bits effective
x86: capped at 0xFF (8 bits), 5 (x86_64) or 6 (ia32) bits effective
s390: capped at 0xFF (8 bits), undocumented effective entropy
Current discussion has led to just dropping the original per-architecture
filters. The additional entropy appears to be safe for arm64, x86,
and s390. Quoting Arnd, "There is no point pretending that 15.75KB is
somehow safe to use while 15.00KB is not."
In some cases specifying the '-n' command line argument will cause
turbostat to fail. For instance 'turbostat -n 1' works fine; however,
'turbostat -n 1 -d' will fail. This is the result of the first call
to getopt_long_only() where "MP" is specified as the optstring. This can
be easily fixed by changing the optstring from "MP" to "MPn:" to remove
ambiguity between the arguments.
tools/power turbostat: option '-n' is ambiguous; possibilities: '-num_iterations' '-no-msr' '-no-perf'
Fixes: a0e86c90b83c ("tools/power turbostat: Add --no-perf option") Signed-off-by: David Arcari <darcari@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
linereq_set_config() behaves badly when direction is not set.
The configuration validation is borrowed from linereq_create(), where,
to verify the intent of the user, the direction must be set to in order to
effect a change to the electrical configuration of a line. But, when
applied to reconfiguration, that validation does not allow for the unset
direction case, making it possible to clear flags set previously without
specifying the line direction.
Adding to the inconsistency, those changes are not immediately applied by
linereq_set_config(), but will take effect when the line value is next get
or set.
For example, by requesting a configuration with no flags set, an output
line with GPIO_V2_LINE_FLAG_ACTIVE_LOW and GPIO_V2_LINE_FLAG_OPEN_DRAIN
set could have those flags cleared, inverting the sense of the line and
changing the line drive to push-pull on the next line value set.
Skip the reconfiguration of lines for which the direction is not set, and
only reconfigure the lines for which direction is set.
Fixes: a54756cb24ea ("gpiolib: cdev: support GPIO_V2_LINE_SET_CONFIG_IOCTL") Signed-off-by: Kent Gibson <warthog618@gmail.com> Link: https://lore.kernel.org/r/20240626052925.174272-3-warthog618@gmail.com Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
linehandle_set_config() behaves badly when direction is not set.
The configuration validation is borrowed from linehandle_create(), where,
to verify the intent of the user, the direction must be set to in order
to effect a change to the electrical configuration of a line. But, when
applied to reconfiguration, that validation does not allow for the unset
direction case, making it possible to clear flags set previously without
specifying the line direction.
Adding to the inconsistency, those changes are not immediately applied by
linehandle_set_config(), but will take effect when the line value is next
get or set.
For example, by requesting a configuration with no flags set, an output
line with GPIOHANDLE_REQUEST_ACTIVE_LOW and GPIOHANDLE_REQUEST_OPEN_DRAIN
requested could have those flags cleared, inverting the sense of the line
and changing the line drive to push-pull on the next line value set.
Ensure the intent of the user by disallowing configurations which do not
have direction set, returning an error to userspace to indicate that the
configuration is invalid.
And, for clarity, use lflags, a local copy of gcnf.flags, throughout when
dealing with the requested flags, rather than a mixture of both.
Fixes: e588bb1eae31 ("gpio: add new SET_CONFIG ioctl() to gpio chardev") Signed-off-by: Kent Gibson <warthog618@gmail.com> Link: https://lore.kernel.org/r/20240626052925.174272-2-warthog618@gmail.com Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Current code configures GCR3 even when device is attached to identity
domain. So that we can support SVA with identity domain. This means in
attach device path it updates Guest Translation related bits in DTE.
Commit de111f6b4f6a ("iommu/amd: Enable Guest Translation after reading
IOMMU feature register") missed to enable Control[GT] bit in resume
path. Its causing certain laptop to fail to resume after suspend.
This is because we have inconsistency between between control register
(GT is disabled) and DTE (where we have enabled guest translation related
bits) in resume path. And IOMMU hardware throws ILLEGAL_DEV_TABLE_ENTRY.
Commit 87a6f1f22c97 ("iommu/amd: Introduce per-device domain ID to fix
potential TLB aliasing issue") introduced per device domain ID when
domain is configured with v2 page table. And in invalidation path, it
uses per device structure (dev_data->gcr3_info.domid) to get the domain ID.
In detach_device() path, current code tries to invalidate IOMMU cache
after removing dev_data from domain device list. This means when domain
is configured with v2 page table, amd_iommu_domain_flush_all() will not be
able to invalidate cache as device is already removed from domain device
list.
This is causing change domain tests (changing domain type from identity to DMA)
to fail with IO_PAGE_FAULT issue.
Hence invalidate cache and update DTE before updating data structures.
Consolidate per device update and flush logic into separate function.
Also make it as global function as it will be used in subsequent series
to update the DTE.
Signed-off-by: Vasant Hegde <vasant.hegde@amd.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/20240418103400.6229-3-vasant.hegde@amd.com Signed-off-by: Joerg Roedel <jroedel@suse.de>
Stable-dep-of: c362f32a59a8 ("iommu/amd: Invalidate cache before removing device from domain list") Signed-off-by: Sasha Levin <sashal@kernel.org>
arch_stack_walk() is called intensively in function_graph when the
kernel is compiled with CONFIG_TRACE_IRQFLAGS. As a result, the kernel
logs a lot of arch_stack_walk and its sub-functions into the ftrace
buffer. However, these functions should not appear on the trace log
because they are part of the ftrace itself. This patch references what
arm64 does for the smae function. So it further prevent the re-enter
kprobe issue, which is also possible on riscv.
Value of pdata->gpio_unbanked is taken from Device Tree. In case of broken
DT due to any error this value can be any. Without this value validation
there can be out of chips->irqs array boundaries access in
davinci_gpio_probe().
Validate the obtained nirq value so that it won't exceed the maximum
number of IRQs per bank.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: eb3744a2dd01 ("gpio: davinci: Do not assume continuous IRQ numbering") Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru> Link: https://lore.kernel.org/r/20240618144344.16943-1-amishin@t-argos.ru Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
KOE TX26D202VM0BWA panel spec indicates the DE signal is active high in
timing chart, so add DISPLAY_FLAGS_DE_HIGH flag in display timing flags.
This aligns display_timing with panel_desc.
The RDMA transport defines values for TSAS, but it cannot be changed as
we only support the 'connected' mode.
So to avoid errors during reconfiguration we should allow to write the
current value.
Fixes: 3f123494db72 ("nvmet: make TCP sectype settable via configfs") Signed-off-by: Hannes Reinecke <hare@kernel.org> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The 'TSAS' value is only defined for TCP and RDMA, but returning
'reserved' for undefined values tricked nvmetcli to try to write
'reserved' when restoring from a config file. This caused an error
and the configuration would not be applied.
Fixes: 3f123494db72 ("nvmet: make TCP sectype settable via configfs") Signed-off-by: Hannes Reinecke <hare@kernel.org> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com> Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
We may leave pat.ops unset when running on brand new platform or
when running as a VF. While the former is unlikely, the latter
is valid (future) use case and will cause NPD when someone will
try to dump PAT settings by debugfs.
It's better to check pointer to pat.ops instead of specific .dump
hook, as we have this hook always defined for every .ops variant.
This is an effort to get rid of all multiplications from allocation
functions in order to prevent integer overflows [1] [2].
In this case, the memory allocated to store RADEONFB_CONN_LIMIT pointers
to "drm_connector" structures can be avoided. This is because this
memory area is never accessed.
Also, in the kzalloc function, it is preferred to use sizeof(*pointer)
instead of sizeof(type) due to the type of the variable can change and
one needs not change the former (unlike the latter).
At the same time take advantage to remove the "#if 0" block, the code
where the removed memory area was accessed, and the RADEONFB_CONN_LIMIT
constant due to now is never used.
Unsupported filesystems currently do not enforce any signatures. Add
support for signature enforcement of the "original" and "portable &
immutable" signatures when EVM_INIT_X509 is enabled.
The "original" signature type contains filesystem specific metadata.
Thus it cannot be copied up and verified. However with EVM_INIT_X509
and EVM_ALLOW_METADATA_WRITES enabled, the "original" file signature
may be written.
When EVM_ALLOW_METADATA_WRITES is not set or once it is removed from
/sys/kernel/security/evm by setting EVM_INIT_HMAC for example, it is not
possible to write or remove xattrs on the overlay filesystem.
This change still prevents EVM from writing HMAC signatures on
unsupported filesystem when EVM_INIT_HMAC is enabled.
In gfs2_jindex_free(), set sdp->sd_jdesc to NULL under the log flush
lock to provide exclusion against gfs2_log_flush().
In gfs2_log_flush(), check if sdp->sd_jdesc is non-NULL before
dereferencing it. Otherwise, we could run into a NULL pointer
dereference when outstanding glock work races with an unmount
(glock_work_func -> run_queue -> do_xmote -> inode_go_sync ->
gfs2_log_flush).
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
mbox_send_message() sends a u32 bit message, not a pointer to a message.
We only convert to a pointer type as a generic type. If we want to send
a dummy message of 0, then simply send 0 (NULL).
An interrupt's effective affinity can only be different from its configured
affinity if there are multiple CPUs. Make it clear that this option is only
meaningful when SMP is enabled. Otherwise, there exists "WARNING: unmet
direct dependencies detected for GENERIC_IRQ_EFFECTIVE_AFF_MASK" when make
menuconfig if CONFIG_SMP is not set on LoongArch.
Because the size passed to copy_from_user() cannot be known beforehand,
it needs to be checked during runtime with check_object_size. That makes
gcc believe that the content of sbuf can be used before init.
Fix:
./include/linux/thread_info.h:215:17: warning: ‘sbuf’ may be used uninitialized [-Wmaybe-uninitialized]
private_key is overwritten with the key parameter passed in by the
caller (if present), or alternatively a newly generated private key.
However, it is possible that the caller provides a key (or the newly
generated key) which is shorter than the previous key. In that
scenario, some key material from the previous key would not be
overwritten. The easiest solution is to explicitly zeroize the entire
private_key array first.
Note that this patch slightly changes the behavior of this function:
previously, if the ecc_gen_privkey failed, the old private_key would
remain. Now, the private_key is always zeroized. This behavior is
consistent with the case where params.key is set and ecc_is_key_valid
fails.
Signed-off-by: Joachim Vandersmissen <git@jvdsn.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sashal@kernel.org>
After firmware boots, it reads keys info from efuse and checks secure
checksum, but suddenly failed to access efuse resulting in probe failure,
and driver throws messages:
For CONFIG_CPUMASK_OFFSTACK=y kernel, explicit allocation of cpumask
variable on stack is not recommended since it can cause potential stack
overflow.
Instead, kernel code should always use *cpumask_var API(s) to allocate
cpumask var in config-neutral way, leaving allocation strategy to
CONFIG_CPUMASK_OFFSTACK.
For CONFIG_CPUMASK_OFFSTACK=y kernel, explicit allocation of cpumask
variable on stack is not recommended since it can cause potential stack
overflow.
Instead, kernel code should always use *cpumask_var API(s) to allocate
cpumask var in config-neutral way, leaving allocation strategy to
CONFIG_CPUMASK_OFFSTACK.
struct rdma_restrack_entry's kern_name was set to KBUILD_MODNAME
in ib_create_cq(), while if the module exited but forgot del this
rdma_restrack_entry, it would cause a invalid address access in
rdma_restrack_clean() when print the owner of this rdma_restrack_entry.
These code is used to help find one forgotten PD release in one of the
ULPs. But it is not needed anymore, so delete them.
It is due to the "void *key" (r2) passed to the helper. bpf allows uninit
stack memory access for bpf prog with the right privileges. This patch
uses kmsan_unpoison_memory() to mark the stack as initialized.
This should address different syzbot reports on the uninit "void *key"
argument during map_{lookup,delete}_elem.
[why]
Driver crashes when pipe idx not set properly
[how]
Add code to skip the pipe that idx not set properly
Reviewed-by: Charlene Liu <charlene.liu@amd.com> Acked-by: Tom Chung <chiahsuan.chung@amd.com> Signed-off-by: Muhammad Ahmed <ahmed.ahmed@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The value of an arithmetic expression directory * master->erasesize is
subject to overflow due to a failure to cast operands to a larger data
type before perfroming arithmetic
Found by Linux Verification Center (linuxtesting.org) with SVACE.
[Why]
Hostvm should be enabled/disabled accordding to the status of
riommu_active, but hostvm always be disabled on DCN31 which causes
underflow
[How]
Set correct hostvm flag on DCN31
Acked-by: Wayne Lin <wayne.lin@amd.com> Signed-off-by: Sherry Wang <Yao.Wang1@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The assembly snippet in restore_fpregs_from_fpstate() that implements
X86_BUG_FXSAVE_LEAK fixup loads the value from a random variable,
preferably the one that is already in the L1 cache.
However, the access to fpinit_state via *fpstate pointer is not
implemented correctly. The "m" asm constraint requires dereferenced
pointer variable, otherwise the compiler just reloads the value
via temporary stack slot. The current asm code reflects this:
mov %rdi,(%rsp)
...
fildl (%rsp)
With dereferenced pointer variable, the code does what the
comment above the asm snippet says:
fildl (%rdi)
Also, remove the pointless %P operand modifier. The modifier is
ineffective on non-symbolic references - it was used to prevent
%rip-relative addresses in .altinstr sections, but FILDL in the
.text section can use %rip-relative addresses without problems.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: https://lore.kernel.org/r/20240315081849.5187-1-ubizjak@gmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Virtio-net driver control queue implementation is not safe
when used with VDUSE. If the VDUSE application does not
reply to control queue messages, it currently ends up
hanging the kernel thread sending this command.
Some work is on-going to make the control queue
implementation robust with VDUSE. Until it is completed,
let's fail features check if control-queue feature is
requested.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Message-Id: <20240109111025.1320976-3-maxime.coquelin@redhat.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Acked-by: Eugenio Pérez <eperezma@redhat.com> Reviewed-by: Xie Yongji <xieyongji@bytedance.com> Acked-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ilitek-ili9881c controls the reset GPIO using the non-sleeping
gpiod_set_value() function. This complains loudly when the GPIO
controller needs to sleep. As the caller can sleep, use
gpiod_set_value_cansleep() to fix the issue.
When auxiliary_device_add() returns error and then calls
auxiliary_device_uninit(), callback function adev_release
calls kfree(madev). We shouldn't call kfree(madev) again
in the error handling path. Set 'madev' to NULL.
Fixes: a69839d4327d ("net: mana: Add support for auxiliary device") Signed-off-by: Ma Ke <make24@iscas.ac.cn> Link: https://patch.msgid.link/20240625130314.2661257-1-make24@iscas.ac.cn Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Even if OOB data is recv()ed, ioctl(SIOCATMARK) must return 1 when the
OOB skb is at the head of the receive queue and no new OOB data is queued.
Without fix:
# RUN msg_oob.no_peek.oob ...
# msg_oob.c:305:oob:Expected answ[0] (0) == oob_head (1)
# oob: Test terminated by assertion
# FAIL msg_oob.no_peek.oob
not ok 2 msg_oob.no_peek.oob
With fix:
# RUN msg_oob.no_peek.oob ...
# OK msg_oob.no_peek.oob
ok 2 msg_oob.no_peek.oob
Now the consumed OOB skb stays at the head of recvq to return a correct
value for ioctl(SIOCATMARK), which is broken now and fixed by a later
patch.
Then, if peer issues recv() with MSG_DONTWAIT, manage_oob() returns NULL,
so recv() ends up with -EAGAIN.
>>> c2.setblocking(False) # This causes -EAGAIN even with available data
>>> c2.recv(5)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
BlockingIOError: [Errno 11] Resource temporarily unavailable
However, next recv() will return the following available data, "world".
>>> c2.recv(5)
b'world'
When the consumed OOB skb is at the head of the queue, we need to fetch
the next skb to fix the weird behaviour.
Note that the issue does not happen without MSG_DONTWAIT because we can
retry after manage_oob().
This patch also adds a test case that covers the issue.
Without fix:
# RUN msg_oob.no_peek.ex_oob_break ...
# msg_oob.c:134:ex_oob_break:AF_UNIX :Resource temporarily unavailable
# msg_oob.c:135:ex_oob_break:Expected:ld
# msg_oob.c:137:ex_oob_break:Expected ret[0] (-1) == expected_len (2)
# ex_oob_break: Test terminated by assertion
# FAIL msg_oob.no_peek.ex_oob_break
not ok 8 msg_oob.no_peek.ex_oob_break
With fix:
# RUN msg_oob.no_peek.ex_oob_break ...
# OK msg_oob.no_peek.ex_oob_break
ok 8 msg_oob.no_peek.ex_oob_break
register store validation for NFT_DATA_VALUE is conditional, however,
the datatype is always either NFT_DATA_VALUE or NFT_DATA_VERDICT. This
only requires a new helper function to infer the register type from the
set datatype so this conditional check can be removed. Otherwise,
pointer to chain object can be leaked through the registers.
The conversion of SPP to MIDI2 UMP called a wrong function, and the
secondary argument wasn't taken. As a result, MSB of SPP was always
zero. Fix to call the right function.
Testing determined that the recent commit 9e046bb111f1 ("tcp: clear
tp->retrans_stamp in tcp_rcv_fastopen_synack()") has a race, and does
not always ensure retrans_stamp is 0 after a TFO payload retransmit.
If transmit completion for the SYN+data skb happens after the client
TCP stack receives the SYNACK (which sometimes happens), then
retrans_stamp can erroneously remain non-zero for the lifetime of the
connection, causing a premature ETIMEDOUT later.
Testing and tracing showed that the buggy scenario is the following
somewhat tricky sequence:
+ Client attempts a TFO handshake. tcp_send_syn_data() sends SYN + TFO
cookie + data in a single packet in the syn_data skb. It hands the
syn_data skb to tcp_transmit_skb(), which makes a clone. Crucially,
it then reuses the same original (non-clone) syn_data skb,
transforming it by advancing the seq by one byte and removing the
FIN bit, and enques the resulting payload-only skb in the
sk->tcp_rtx_queue.
+ Client sets retrans_stamp to the start time of the three-way
handshake.
+ Cookie mismatches or server has TFO disabled, and server only ACKs
SYN.
+ tcp_ack() sees SYN is acked, tcp_clean_rtx_queue() clears
retrans_stamp.
+ Since the client SYN was acked but not the payload, the TFO failure
code path in tcp_rcv_fastopen_synack() tries to retransmit the
payload skb. However, in some cases the transmit completion for the
clone of the syn_data (which had SYN + TFO cookie + data) hasn't
happened. In those cases, skb_still_in_host_queue() returns true
for the retransmitted TFO payload, because the clone of the syn_data
skb has not had its tx completetion.
+ Because skb_still_in_host_queue() finds skb_fclone_busy() is true,
it sets the TSQ_THROTTLED bit and the retransmit does not happen in
the tcp_rcv_fastopen_synack() call chain.
+ The tcp_rcv_fastopen_synack() code next implicitly assumes the
retransmit process is finished, and sets retrans_stamp to 0 to clear
it, but this is later overwritten (see below).
+ Later, upon tx completion, tcp_tsq_write() calls
tcp_xmit_retransmit_queue(), which puts the retransmit in flight and
sets retrans_stamp to a non-zero value.
+ The client receives an ACK for the retransmitted TFO payload data.
+ Since we're in CA_Open and there are no dupacks/SACKs/DSACKs/ECN to
make tcp_ack_is_dubious() true and make us call
tcp_fastretrans_alert() and reach a code path that clears
retrans_stamp, retrans_stamp stays nonzero.
+ Later, if there is a TLP, RTO, RTO sequence, then the connection
will suffer an early ETIMEDOUT due to the erroneously ancient
retrans_stamp.
The fix: this commit refactors the code to have
tcp_rcv_fastopen_synack() retransmit by reusing the relevant parts of
tcp_simple_retransmit() that enter CA_Loss (without changing cwnd) and
call tcp_xmit_retransmit_queue(). We have tcp_simple_retransmit() and
tcp_rcv_fastopen_synack() share code in this way because in both cases
we get a packet indicating non-congestion loss (MTU reduction or TFO
failure) and thus in both cases we want to retransmit as many packets
as cwnd allows, without reducing cwnd. And given that retransmits will
set retrans_stamp to a non-zero value (and may do so in a later
calling context due to TSQ), we also want to enter CA_Loss so that we
track when all retransmitted packets are ACked and clear retrans_stamp
when that happens (to ensure later recurring RTOs are using the
correct retrans_stamp and don't declare ETIMEDOUT prematurely).
Fixes: 9e046bb111f1 ("tcp: clear tp->retrans_stamp in tcp_rcv_fastopen_synack()") Fixes: a7abf3cd76e1 ("tcp: consider using standard rtx logic in tcp_rcv_fastopen_synack()") Signed-off-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Link: https://patch.msgid.link/20240624144323.2371403-1-ncardwell.sw@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If we're not in a NAPI softirq context, we need to be careful
about how we call napi_consume_skb(), specifically we need to
call it with budget==0 to signal to it that we're not in a
safe context.
This was found while running some configuration stress testing
of traffic and a change queue config loop running, and this
curious note popped out:
I found that ionic_tx_clean() calls napi_consume_skb() which calls
napi_skb_cache_put(), but before that last call is the note
/* Zero budget indicate non-NAPI context called us, like netpoll */
and
DEBUG_NET_WARN_ON_ONCE(!in_softirq());
Those are pretty big hints that we're doing it wrong. We can pass a
context hint down through the calls to let ionic_tx_clean() know what
we're doing so it can call napi_consume_skb() correctly.
Fixes: 386e69865311 ("ionic: Make use napi_consume_skb") Signed-off-by: Shannon Nelson <shannon.nelson@amd.com> Link: https://patch.msgid.link/20240624175015.4520-1-shannon.nelson@amd.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Johannes missed parisc back when he introduced the compat version
of these syscalls, so receiving cmsg messages that require a compat
conversion is still broken.
Use the correct calls like the other architectures do.
sparc has the wrong compat version of recv() and recvfrom() for both the
direct syscalls and socketcall().
The direct syscalls just need to use the compat version. For socketcall,
the same thing could be done, but it seems better to completely remove
the custom assembler code for it and just use the same implementation that
everyone else has.
Fixes: 1dacc76d0014 ("net/compat/wext: send different messages to compat tasks") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
sparc has two identical select syscalls at numbers 93 and 230, respectively.
During the conversion to the modern syscall.tbl format, the older one of the
two broke in compat mode, and now refers to the native 64-bit syscall.
Restore the correct behavior. This has very little effect, as glibc has
been using the newer number anyway.
The switch global port interrupt mask, REG_SW_PORT_INT_MASK__4, is
defined as 0x001C in ksz9477_reg.h. The designers used 32-bit value in
anticipation for increase of port count in future product but currently
the maximum port count is 7 and the effective value is 0x7F in register
0x001F. Each port has its own interrupt mask and is defined as 0x#01F.
It uses only 4 bits for different interrupts.
The developer who implemented the current interrupt mechanism in the
switch driver noticed there are similarities between the mechanism to
mask port interrupts in global interrupt and individual interrupts in
each port and so used the same code to handle these interrupts. He
updated the code to use the new macro REG_SW_PORT_INT_MASK__1 which is
defined as 0x1F in ksz_common.h but he forgot to update the 32-bit write
to 8-bit as now the mask registers are 0x1F and 0x#01F.
In addition all KSZ switches other than the KSZ9897/KSZ9893 and LAN937X
families use only 8-bit access and so this common code will eventually
be changed to accommodate them.
projects / thirdparty / kernel / stable.git / log
