The pinctrl->functions[] array has pinctrl->num_functions elements and
the pinctrl->groups[] array is the same way. These are set in
ns2_pinmux_probe(). So the > comparisons should be >= so that we don't
read one element beyond the end of the array.
Fixes: b5aa1006e4 ("pinctrl: ns2: add pinmux driver support for Broadcom NS2 SoC")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/20190926081426.GB2332@mwanda
Acked-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This convert the BCM IPROC driver to use the SPDX header
for indicating GPL v2.0 only licensing.
Cc: Pramod Kumar <pramodku@broadcom.com>
Cc: Ray Jui <rjui@broadcom.com>
Cc: Scott Branden <sbranden@broadcom.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Scott Branden <scott.branden@broadcom.com>
Link: https://lore.kernel.org/r/20191002130217.4491-1-linus.walleij@linaro.org
The GPIO controller doesn't have any controls to enable the system to
wake up from low power states based on activity on GPIO pins. An extra
hardware block that is part of the power management controller (PMC)
contains these controls. In order for the GPIO controller to be able
to cooperate with the PMC, obtain a reference to the PMC's IRQ domain
and make it a parent to the GPIO controller's IRQ domain. This way the
PMC gets an opportunity to program the additional registers required
to enable wakeup sources on suspend.
Based on additional work by Bitan Biswas <bbiswas@nvidia.com>.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20191002144502.156393-2-thierry.reding@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The interrupt-related register fields on the MAX77620 GPIO controller
share registers with GPIO related fields. If the IRQ chip is implemented
with regmap-irq, this causes the IRQ controller code to overwrite fields
previously configured by the GPIO controller code.
Two examples where this causes problems are the NVIDIA Jetson TX1 and
Jetson TX2 boards, where some of the GPIOs are used to enable vital
power regulators. The MAX77620 GPIO controller also provides the USB OTG
ID pin. If configured as an interrupt, this causes some of the
regulators to be powered off.
Signed-off-by: Timo Alho <talho@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20191002122825.3948322-3-thierry.reding@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
regmap_add_irq_chip() will try to allocate all of the IRQ descriptors
upfront if passed a non-zero irq_base parameter. However, the intention
is to allocate IRQ descriptors on an as-needed basis if possible. Pass 0
instead of -1 to fix that use-case.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20191002122825.3948322-2-thierry.reding@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The gpiod_set_debounce() function takes the debounce time in
microseconds. Adjust the switch/case values in the MAX77620 GPIO to use
the correct unit.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20191002122825.3948322-1-thierry.reding@gmail.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The 37xx configuration registers are only 32 bits long, so
pins 32-35 spill over into the next register. The calculation
for the register address was done, but the bitmask was not, so
any configuration to pin 32 or above resulted in a bitmask that
overflowed and performed no action.
Fix the register / offset calculation to also adjust the offset.
Fixes: 5715092a45 ("pinctrl: armada-37xx: Add gpio support")
Signed-off-by: Patrick Williams <alpawi@amazon.com>
Acked-by: Gregory CLEMENT <gregory.clement@bootlin.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20191001154634.96165-1-alpawi@amazon.com
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Amazingly, of all features, this does not require a switch reset.
Tested with:
tc qdisc add dev swp2 clsact
tc filter add dev swp2 ingress matchall skip_sw \
action mirred egress mirror dev swp3
tc filter show dev swp2 ingress
tc filter del dev swp2 ingress pref 49152
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When fetching free MSI-X vectors for ULDs, check for the error code
before accessing MSI-X info array. Otherwise, an out-of-bounds access is
attempted, which results in kernel panic.
Fixes: 94cdb8bb99 ("cxgb4: Add support for dynamic allocation of resources for ULD")
Signed-off-by: Shahjada Abul Husain <shahjada@chelsio.com>
Signed-off-by: Vishal Kulkarni <vishal@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This reverts commit a3ce2a21bb.
Eric reported tests failings with commit. After digging into it,
the bottom line is that the DAD sequence is not to be messed with.
There are too many cases that are expected to proceed regardless
of whether a device is up.
Revert the patch and I will send a different solution for the
problem Rajendra reported.
Signed-off-by: David Ahern <dsahern@gmail.com>
Cc: Eric Dumazet <edumazet@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This function returns string literals which are "const char *".
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This function is only used via function pointer.
"inline" doesn't hurt given that taking address of an inline function
forces out-of-line version but it doesn't help either.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The helper pinctrl_dt_has_hogs() was introduced in
99e4f67508 (pinctrl: core: Use delayed work for hogs), but the sole
use then got removed shortly after in 950b0d91dc (pinctrl: core: Fix
regression caused by delayed work for hogs).
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Link: https://lore.kernel.org/r/20190923142005.5632-1-linux@rasmusvillemoes.dk
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Add casts to fix these warnings:
./usr/include/linux/netfilter_arp/arp_tables.h:200:19: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/netfilter_bridge/ebtables.h:197:19: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/netfilter_ipv4/ip_tables.h:223:19: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/netfilter_ipv6/ip6_tables.h:263:19: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/tipc_config.h:310:28: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/tipc_config.h:410:24: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
./usr/include/linux/virtio_ring.h:170:16: error: pointer of type 'void *' used in arithmetic [-Werror=pointer-arith]
Those are theoretical probably but kernel doesn't control compiler flags
in userspace.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a comment about why the call to of_match_node() cannot be replaced
by of_device_get_match_data(). This will hopefully prevent people from
attempting to clean this up in the future.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Link: https://lore.kernel.org/r/20190923095400.GA11084@ulmo
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Florian Fainelli says:
====================
net: phy: broadcom: RGMII delays fixes
This patch series fixes the BCM54210E RGMII delay configuration which
could only have worked in a PHY_INTERFACE_MODE_RGMII configuration.
There is a forward declaration added such that the first patch can be
picked up for -stable and apply fine all the way back to when the bug
was introduced.
The second patch eliminates duplicated code that used a different kind
of logic and did not use existing constants defined.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
bcm54612e_config_init() duplicates what bcm54xx_config_clock_delay()
does with respect to configuring RGMII TX/RX delays appropriately.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 0fc9ae1076 ("net: phy: broadcom: add support for
BCM54210E") added support for BCM54210E but also unconditionally cleared
the RXC to RXD skew and the TXD to TXC skew, thus only making
PHY_INTERFACE_MODE_RGMII a possible configuration. Use
bcm54xx_config_clock_delay() which correctly sets the registers
depending on the 4 possible PHY interface values that exist for RGMII.
Fixes: 0fc9ae1076 ("net: phy: broadcom: add support for BCM54210E")
Reported-by: Manasa Mudireddy <manasa.mudireddy@broadcom.com>
Reported-by: Ray Jui <ray.jui@broadcom.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The call to pinctrl_count_index_with_args checks for a -EINVAL return
however this function calls pinctrl_get_list_and_count and this can
return -ENOENT. Rather than check for a specific error, fix this by
checking for any error return to catch the -ENOENT case.
Addresses-Coverity: ("Improper use of negative")
Fixes: 003910ebc8 ("pinctrl: Introduce TI IOdelay configuration driver")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20190920122030.14340-1-colin.king@canonical.com
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Jakub Kicinski says:
====================
net/tls: separate the TLS TOE code out
We have 3 modes of operation of TLS - software, crypto offload
(Mellanox, Netronome) and TCP Offload Engine-based (Chelsio).
The last one takes over the socket, like any TOE would, and
is not really compatible with how we want to do things in the
networking stack.
Confusingly the name of the crypto-only offload mode is TLS_HW,
while TOE-offload related functions use tls_hw_ as their prefix.
Engineers looking to implement offload are also be faced with
TOE artefacts like struct tls_device (while, again,
CONFIG_TLS_DEVICE actually gates the non-TOE offload).
To improve the clarity of the offload code move the TOE code
into new files, and rename the functions and structures
appropriately.
Because TOE-offload takes over the socket, and makes no use of
the TLS infrastructure in the kernel, the rest of the code
(anything beyond the ULP setup handlers) do not have to worry
about the mode == TLS_HW_RECORD case.
The increase in code size is due to duplication of the full
license boilerplate. Unfortunately original author (Dave Watson)
seems unreachable :(
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
TLS "record layer offload" requires TOE, and bypasses most of
the normal networking stack. It is also significantly less
maintained. Allow users to compile it out to avoid issues.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The tls_hw_* functions are quite confusingly named, since they
are related to the TOE-offload, not TLS_HW offload which doesn't
require TOE. Rename them.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move tls_hw_* functions to a new, separate source file
to avoid confusion with normal, non-TOE offload.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move tls_build_proto() so that TOE offload doesn't have to call it
mid way through its bypass enable path.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename struct tls_device to struct tls_toe_device to avoid
confusion with normal, non-TOE offload.
No functional changes.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move tls_device structure and register/unregister functions
to a new header to avoid confusion with normal, non-TOE offload.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There was supposed to be a trace indicating that a new peer had been
created. Add it.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the rxrpc_recvmsg tracepoint to handle being called with a NULL call
parameter.
Fixes: a25e21f0bc ("rxrpc, afs: Use debug_ids rather than pointers in traces")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A sysfs request to enable or disable a PCIe hotplug slot should not
return before it has been carried out. That is sought to be achieved by
waiting until the controller's "pending_events" have been cleared.
However the IRQ thread pciehp_ist() clears the "pending_events" before
it acts on them. If pciehp_sysfs_enable_slot() / _disable_slot() happen
to check the "pending_events" after they have been cleared but while
pciehp_ist() is still running, the functions may return prematurely
with an incorrect return value.
Fix by introducing an "ist_running" flag which must be false before a sysfs
request is allowed to return.
Fixes: 32a8cef274 ("PCI: pciehp: Enable/disable exclusively from IRQ thread")
Link: https://lore.kernel.org/linux-pci/1562226638-54134-1-git-send-email-wangxiongfeng2@huawei.com
Link: https://lore.kernel.org/r/4174210466e27eb7e2243dd1d801d5f75baaffd8.1565345211.git.lukas@wunner.de
Reported-and-tested-by: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v4.19+
generic/018 reports an inconsistent status of atime, the
testcase is as below:
- open file with O_SYNC
- write file to construct fraged space
- calc md5 of file
- record {a,c,m}time
- defrag file --- do nothing
- umount & mount
- check {a,c,m}time
The root cause is, as f2fs enables lazytime by default, atime
update will dirty vfs inode, rather than dirtying f2fs inode (by set
with FI_DIRTY_INODE), so later f2fs_write_inode() called from VFS will
fail to update inode page due to our skip:
f2fs_write_inode()
if (is_inode_flag_set(inode, FI_DIRTY_INODE))
return 0;
So eventually, after evict(), we lose last atime for ever.
To fix this issue, we need to check whether {a,c,m,cr}time is
consistent in between inode cache and inode page, and only skip
f2fs_update_inode() if f2fs inode is not dirty and time is
consistent as well.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
- Build fixes for Cavium Octeon & PMC-Sierra MSP systems, as well as
all pre-MIPSr6 configurations built with binutils < 2.25.
- Boot fixes for 64-bit Loongson systems & SGI IP28 systems.
- Wire up the new clone3 syscall.
- Clean ups for a few build-time warnings.
-----BEGIN PGP SIGNATURE-----
iIsEABYIADMWIQRgLjeFAZEXQzy86/s+p5+stXUA3QUCXZekBBUccGF1bC5idXJ0
b25AbWlwcy5jb20ACgkQPqefrLV1AN1aSQD+OSdb9fs1Gqakn1bM/TmvAB5lNxhM
8Xy47aqWURt4rpIBAJegfqjq2SWyVNxeNV0fNcqQl9DcjABhOyCJ316U6rMD
=TNtT
-----END PGP SIGNATURE-----
Merge tag 'mips_fixes_5.4_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux
Pull MIPS fixes from Paul Burton:
- Build fixes for Cavium Octeon & PMC-Sierra MSP systems, as well as
all pre-MIPSr6 configurations built with binutils < 2.25.
- Boot fixes for 64-bit Loongson systems & SGI IP28 systems.
- Wire up the new clone3 syscall.
- Clean ups for a few build-time warnings.
* tag 'mips_fixes_5.4_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
MIPS: fw/arc: Remove unused addr variable
MIPS: pmcs-msp71xx: Remove unused addr variable
MIPS: pmcs-msp71xx: Add missing MAX_PROM_MEM definition
mips: Loongson: Fix the link time qualifier of 'serial_exit()'
MIPS: init: Prevent adding memory before PHYS_OFFSET
MIPS: init: Fix reservation of memory between PHYS_OFFSET and mem start
MIPS: VDSO: Fix build for binutils < 2.25
MIPS: VDSO: Remove unused gettimeofday.c
MIPS: Wire up clone3 syscall
MIPS: octeon: Include required header; fix octeon ethernet build
MIPS: cpu-bugs64: Mark inline functions as __always_inline
MIPS: dts: ar9331: fix interrupt-controller size
MIPS: Loongson64: Fix boot failure after dropping boot_mem_map
Two RISC-V fixes for v5.4-rc2:
- Ensure that exclusive-load reservations are terminated after system
call or exception handling. This primarily affects QEMU, which does
not expire load reservations.
- Fix an issue primarily affecting RV32 platforms that can cause the
DT header to be corrupted, causing boot failures.
-----BEGIN PGP SIGNATURE-----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=yka9
-----END PGP SIGNATURE-----
Merge tag 'riscv/for-v5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux
Pull RISC-V fixes from Paul Walmsley:
- Ensure that exclusive-load reservations are terminated after system
call or exception handling. This primarily affects QEMU, which does
not expire load reservations.
- Fix an issue primarily affecting RV32 platforms that can cause the DT
header to be corrupted, causing boot failures.
* tag 'riscv/for-v5.4-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
riscv: Fix memblock reservation for device tree blob
RISC-V: Clear load reservations while restoring hart contexts
The RockPro64 schematic [1] page 18 states a min voltage of 0.8V and a
max voltage of 1.4V for the VDD_LOG pwm regulator. However, there is an
additional note that the pwm parameter needs to be modified.
From the schematics a voltage range of 0.8V to 1.7V can be calculated.
Additional voltage measurements on the board show that this fix indeed
leads to the correct voltage, while without this fix the voltage was set
too high.
[1] http://files.pine64.org/doc/rockpro64/rockpro64_v21-SCH.pdf
Fixes: e4f3fb4909 ("arm64: dts: rockchip: add initial dts support for Rockpro64")
Signed-off-by: Soeren Moch <smoch@web.de>
Link: https://lore.kernel.org/r/20191003215036.15023-1-smoch@web.de
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Now that instances of input_dev support polling mode natively,
we no longer need to create input_polled_dev instance.
Acked-by: Hans-Christian Noren Egtvedt <egtvedt@samfundet.no>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Some drivers need to be able to know the current polling interval for
devices working in polling mode, let's allow them fetching it.
Acked-By: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Tested-by: Michal Vokáč <michal.vokac@ysoft.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Add a loop test with 32 bit register against 0 immediate:
# ./test_verifier 631
#631/p taken loop with back jump to 1st insn, 2 OK
Disassembly:
[...]
1b: test %edi,%edi
1d: jne 0x0000000000000014
[...]
Pretty much similar to prior "taken loop with back jump to 1st
insn" test case just as jmp32 variant.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Replace 'cmp reg, 0' with 'test reg, reg' for comparisons against
zero. Saves 1 byte of instruction encoding per occurrence. The flag
results of test 'reg, reg' are identical to 'cmp reg, 0' in all
cases except for AF which we don't use/care about. In terms of
macro-fusibility in combination with a subsequent conditional jump
instruction, both have the same properties for the jumps used in
the JIT translation. For example, same JITed Cilium program can
shrink a bit from e.g. 12,455 to 12,317 bytes as tests with 0 are
used quite frequently.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
While MR uses live as the SRCU 'update', the MW case uses the xarray
directly, xa_erase() causes the MW to become inaccessible to the pagefault
thread.
Thus whenever a MW is removed from the xarray we must synchronize_srcu()
before freeing it.
This must be done before freeing the mkey as re-use of the mkey while the
pagefault thread is using the stale mkey is undesirable.
Add the missing synchronizes to MW and DEVX indirect mkey and delete the
bogus protection against double destroy in mlx5_core_destroy_mkey()
Fixes: 534fd7aac5 ("IB/mlx5: Manage indirection mkey upon DEVX flow for ODP")
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-7-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
live is used to signal to the pagefault thread that the MR is initialized
and ready for use. It should be after the umem is assigned and all other
setup is completed. This prevents races (at least) of the form:
CPU0 CPU1
mlx5_ib_alloc_implicit_mr()
implicit_mr_alloc()
live = 1
imr->umem = umem
num_pending_prefetch_inc()
if (live)
atomic_inc(num_pending_prefetch)
atomic_set(num_pending_prefetch,0) // Overwrites other thread's store
Further, live is being used with SRCU as the 'update' in an
acquire/release fashion, so it can not be read and written raw.
Move all live = 1's to after MR initialization is completed and use
smp_store_release/smp_load_acquire() for manipulating it.
Add a missing live = 0 when an implicit MR child is deleted, before
queuing work to do synchronize_srcu().
The barriers in update_odp_mr() were some broken attempt to create a
acquire/release, but were not even applied consistently and missed the
point, delete it as well.
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-6-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
During destroy setting live = 0 and then synchronize_srcu() prevents
num_pending_prefetch from incrementing, and also, ensures that all work
holding that count is queued on the WQ. Testing before causes races of the
form:
CPU0 CPU1
dereg_mr()
mlx5_ib_advise_mr_prefetch()
srcu_read_lock()
num_pending_prefetch_inc()
if (!live)
live = 0
atomic_read() == 0
// skip flush_workqueue()
atomic_inc()
queue_work();
srcu_read_unlock()
WARN_ON(atomic_read()) // Fails
Swap the order so that the synchronize_srcu() prevents this.
Fixes: a6bc3875f1 ("IB/mlx5: Protect against prefetch of invalid MR")
Link: https://lore.kernel.org/r/20191001153821.23621-5-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This fixes a race of the form:
CPU0 CPU1
mlx5_ib_invalidate_range() mlx5_ib_invalidate_range()
// This one actually makes npages == 0
ib_umem_odp_unmap_dma_pages()
if (npages == 0 && !dying)
// This one does nothing
ib_umem_odp_unmap_dma_pages()
if (npages == 0 && !dying)
dying = 1;
dying = 1;
schedule_work(&umem_odp->work);
// Double schedule of the same work
schedule_work(&umem_odp->work); // BOOM
npages and dying must be read and written under the umem_mutex lock.
Since whenever ib_umem_odp_unmap_dma_pages() is called mlx5 must also call
mlx5_ib_update_xlt, and both need to be done in the same locking region,
hoist the lock out of unmap.
This avoids an expensive double critical section in
mlx5_ib_invalidate_range().
Fixes: 81713d3788 ("IB/mlx5: Add implicit MR support")
Link: https://lore.kernel.org/r/20191001153821.23621-4-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
mlx5_ib_update_xlt() must be protected against parallel free of the MR it
is accessing, also it must be called single threaded while updating the
HW. Otherwise we can have races of the form:
CPU0 CPU1
mlx5_ib_update_xlt()
mlx5_odp_populate_klm()
odp_lookup() == NULL
pklm = ZAP
implicit_mr_get_data()
implicit_mr_alloc()
<update interval tree>
mlx5_ib_update_xlt
mlx5_odp_populate_klm()
odp_lookup() != NULL
pklm = VALID
mlx5_ib_post_send_wait()
mlx5_ib_post_send_wait() // Replaces VALID with ZAP
This can be solved by putting both the SRCU and the umem_mutex lock around
every call to mlx5_ib_update_xlt(). This ensures that the content of the
interval tree relavent to mlx5_odp_populate_klm() (ie mr->parent == mr)
will not change while it is running, and thus the posted WRs to update the
KLM will always reflect the correct information.
The race above will resolve by either having CPU1 wait till CPU0 completes
the ZAP or CPU0 will run after the add and instead store VALID.
The pagefault path adding children already holds the umem_mutex and SRCU,
so the only missed lock is during MR destruction.
Fixes: 81713d3788 ("IB/mlx5: Add implicit MR support")
Link: https://lore.kernel.org/r/20191001153821.23621-3-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
This code is completely broken, the umem of a ODP MR simply cannot be
discarded without a lot more locking, nor can an ODP mkey be blithely
destroyed via destroy_mkey().
Fixes: 6aec21f6a8 ("IB/mlx5: Page faults handling infrastructure")
Link: https://lore.kernel.org/r/20191001153821.23621-2-jgg@ziepe.ca
Reviewed-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
rdma_for_each_port is already incrementing the iterator's value it
receives therefore, after the first iteration the iterator is increased by
2 which eventually causing wrong queries and possible traces.
Fix the above by removing the old redundant incrementation that was used
before rdma_for_each_port() macro.
Cc: <stable@vger.kernel.org>
Fixes: ea1075edcb ("RDMA: Add and use rdma_for_each_port")
Link: https://lore.kernel.org/r/20191002122127.17571-1-leon@kernel.org
Signed-off-by: Mohamad Heib <mohamadh@mellanox.com>
Reviewed-by: Erez Alfasi <ereza@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
