This ensures that we return the right structures back to userspace.
Otherwise, it looks like the reserved fields in the response structures
in userspace might have uninitialized data in them.
I was not seeing my linker flags getting added when using ld-option when
cross compiling with Clang. Upon investigation, this seems to be due to
a difference in how GCC vs Clang handle cross compilation.
GCC is configured at build time to support one backend, that is implicit
when compiling. Clang is explicit via the use of `-target <triple>` and
ships with all supported backends by default.
GNU Make feature test macros that compile then link will always fail
when cross compiling with Clang unless Clang's triple is passed along to
the compiler. For example:
$ clang -x c /dev/null -c -o temp.o
$ aarch64-linux-android/bin/ld -E temp.o
aarch64-linux-android/bin/ld:
unknown architecture of input file `temp.o' is incompatible with
aarch64 output
aarch64-linux-android/bin/ld:
warning: cannot find entry symbol _start; defaulting to 0000000000400078
$ echo $?
1
$ clang -target aarch64-linux-android- -x c /dev/null -c -o temp.o
$ aarch64-linux-android/bin/ld -E temp.o
aarch64-linux-android/bin/ld:
warning: cannot find entry symbol _start; defaulting to 00000000004002e4
$ echo $?
0
This causes conditional checks that invoke $(CC) without the target
triple, then $(LD) on the result, to always fail.
Users can provide garbage while calling to ucma_join_ip_multicast(),
it will indirectly cause to rdma_addr_size() return 0, making the
call to ucma_process_join(), which had the right checks, but it is
better to check the input as early as possible.
Fixes: 5bc2b7b397b0 ("RDMA/ucma: Allow user space to specify AF_IB when joining multicast") Reported-by: <syzbot+2287ac532caa81900a4e@syzkaller.appspotmail.com> Signed-off-by: Leon Romanovsky <leonro@mellanox.com> Reviewed-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The orphan clocks reparents should migrate any existing count from the
orphan clock to its new acestor clocks, otherwise we may have
inconsistent counts in the tree and end-up with gated critical clocks
Assuming we have two clocks, A and B.
* Clock A has CLK_IS_CRITICAL flag set.
* Clock B is an ancestor of A which can gate. Clock B gate is left
enabled by the bootloader.
Step 1: Clock A is registered. Since it is a critical clock, it is
enabled. The clock being still an orphan, no parent are enabled.
Step 2: Clock B is registered and reparented to clock A (potentially
through several other clocks). We are now in situation where the enable
count of clock A is 1 while the enable count of its ancestors is 0, which
is not good.
Step 3: in lateinit, clk_disable_unused() is called, the enable_count of
clock B being 0, clock B is gated and and critical clock A actually gets
disabled.
This situation was found while adding fdiv_clk gates to the meson8b
platform. These clocks parent clk81 critical clock, which is the mother
of all peripheral clocks in this system. Because of the issue described
here, the system is crashing when clk_disable_unused() is called.
The situation is solved by reverting
commit f8f8f1d04494 ("clk: Don't touch hardware when reparenting during registration").
To avoid breaking again the situation described in this commit
description, enabling critical clock should be done before walking the
orphan list. This way, a parent critical clock may not be accidentally
disabled due to the CLK_OPS_PARENT_ENABLE mechanism.
Fixes: f8f8f1d04494 ("clk: Don't touch hardware when reparenting during registration") Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Shawn Guo <shawnguo@kernel.org> Cc: Dong Aisheng <aisheng.dong@nxp.com> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com> Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Tested-by: Heiko Stuebner <heiko@sntech.de> Signed-off-by: Michael Turquette <mturquette@baylibre.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
cma_port_is_unique() allows local port reuse if the quad (source
address and port, destination address and port) for this connection
is unique. However, if the destination info is zero or unspecified, it
can't make a correct decision but still allows port reuse. For example,
sometimes rdma_bind_addr() is called with unspecified destination and
reusing the port can lead to creating a connection with a duplicate quad,
after the destination is resolved. The issue manifests when MPI scale-up
tests hang after the duplicate quad is used.
Set the destination address family and add checks for zero destination
address and port to prevent source port reuse based on invalid destination.
Fixes: 19b752a19dce ("IB/cma: Allow port reuse for rdma_id") Reviewed-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com> Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Do not fail on multiport cards in serial_pci_is_class_communication().
It restores behaviour for SUNIX multiport cards, that enumerated by
class and have a custom board data.
Moreover it allows users to reenumerate port-by-port from user space.
Fixes: 7d8905d06405 ("serial: 8250_pci: Enable device after we check black list") Reported-by: Nikola Ciprich <nikola.ciprich@linuxbox.cz> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Tested-by: Nikola Ciprich <nikola.ciprich@linuxbox.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Memory state around the buggy address: ffff880066b99180: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ffff880066b99200: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffff880066b99280: 00 00 fc fc fc fc fc fc fc fc fc fc fc fc fc fc
^ ffff880066b99300: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff880066b99380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
Cc: syzkaller <syzkaller@googlegroups.com> Fixes: 0fb2ed66a14c ("IB/mlx5: Add create and destroy functionality for Raw Packet QP") Signed-off-by: Boris Pismenny <borisp@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Doug Ledford <dledford@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch finishes all outstanding SCSI IO commands (but not other commands,
e.g., task management) in the shutdown and unload paths.
It first waits for the commands to complete (this is done after setting
'ioc->remove_host = 1 ', which prevents new commands to be queued) then it
flushes commands that might still be running.
This avoids triggering error handling (e.g., abort command) for all commands
possibly completed by the adapter after interrupts disabled.
[mauricfo: introduced something in commit message.]
Signed-off-by: Sreekanth Reddy <sreekanth.reddy@broadcom.com> Tested-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
[mauricfo: backport to linux-4.14.y (a few updates to context lines)] Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds checks for 'ioc->remove_host' in the SCSI error handlers, so
not to access pointers/resources potentially freed in the PCI shutdown/module
unload path. The error handlers may be invoked after shutdown/unload,
depending on other components.
This problem was observed with kexec on a system with a mpt3sas based adapter
and an infiniband adapter which takes long enough to shutdown:
The mpt3sas driver finished shutting down / disabled interrupt handling, thus
some commands have not finished and timed out.
Since the system was still running (waiting for the infiniband adapter to
shutdown), the scsi error handler for task abort of mpt3sas was invoked, and
hit an oops -- either in scsih_abort() because 'ioc->scsi_lookup' was NULL
without commit dbec4c9040ed ("scsi: mpt3sas: lockless command submission"), or
later up in scsih_host_reset() (with or without that commit), because it
eventually called mpt3sas_base_get_iocstate().
After the above commit, the oops in scsih_abort() does not occur anymore
(_scsih_scsi_lookup_find_by_scmd() is no longer called), but that commit is
too big and out of the scope of linux-stable, where this patch might help, so
still go for the changes.
Also, this might help to prevent similar errors in the future, in case code
changes and possibly tries to access freed stuff.
Note the fix in scsih_host_reset() is still important anyway.
Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Acked-by: Sreekanth Reddy <Sreekanth.Reddy@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Register layout of a typical TPCC_EVT_MUX_M_N register is such that the
lowest numbered event is at the lowest byte address and highest numbered
event at highest byte address. But TPCC_EVT_MUX_60_63 register layout is
different, in that the lowest numbered event is at the highest address
and highest numbered event is at the lowest address. Therefore, modify
ti_am335x_xbar_write() to handle TPCC_EVT_MUX_60_63 register
accordingly.
Signed-off-by: Vignesh R <vigneshr@ti.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Signed-off-by: Vinod Koul <vinod.koul@intel.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This drivers probe fails due to a clock name collision if a clock named
'plla' or 'pllb' is already registered when registering this drivers
internal plls.
Fix it by renaming internal plls to avoid name collisions.
If the nocount bit is set the divider is bypassed and the settings for the
divider count should be ignored and a divider value of 1 should be assumed.
Handle this correctly in the driver recalc_rate() callback.
While the driver sets up the part so that the read back dividers values
yield the correct result the power-on reset settings of the part might not
reflect this and hence calling e.g. clk_get_rate() without prior calls to
clk_set_rate() will yield the wrong result.
The orphan clocks reparent operation shouldn't touch the hardware
if clocks are enabled, otherwise it may get a chance to disable a
newly registered critical clock which triggers the warning below.
Assuming we have two clocks: A and B, B is the parent of A.
Clock A has flag: CLK_OPS_PARENT_ENABLE
Clock B has flag: CLK_IS_CRITICAL
Step 1:
Clock A is registered, then it becomes orphan.
Step 2:
Clock B is registered. Before clock B reach the critical clock enable
operation, orphan A will find the newly registered parent B and do
reparent operation, then parent B will be finally disabled in
__clk_set_parent_after() due to CLK_OPS_PARENT_ENABLE flag as there's
still no users of B which will then trigger the following warning.
We know that the clk isn't enabled with any sort of prepare_count
here so we don't need to enable anything to prevent a race. And
we're holding the prepare mutex so set_rate/set_parent can't race
here either. Based on an earlier patch by Dong Aisheng.
Fixes: fc8726a2c021 ("clk: core: support clocks which requires parents enable (part 2)") Cc: Michael Turquette <mturquette@baylibre.com> Cc: Shawn Guo <shawnguo@kernel.org> Reported-by: Dong Aisheng <aisheng.dong@nxp.com> Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Wait for the syncronization of all clocks when resuming, not only the
UPLL clock. Do not use regmap_read_poll_timeout, as it will call BUG()
when interrupts are masked, which is the case in here.
Debugfs file reset_stats is created with S_IRUSR permissions,
but ocrdma_dbgfs_ops_read() doesn't support OCRDMA_RESET_STATS,
whereas ocrdma_dbgfs_ops_write() supports only OCRDMA_RESET_STATS.
The patch fixes misstype with permissions.
Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru> Acked-by: Selvin Xavier <selvin.xavier@broadcom.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the lpfc discovery engine, when as a nvme target, where the driver
was performing mailbox io with the adapter for port login when a NVME
PRLI is received from the host. Rather than queue and eventually get
back to sending a response after the mailbox traffic, the driver
rejected the io with an error response.
Turns out this particular initiator didn't like the rejection values
(unable to process command/command in progress) so it never attempted a
retry of the PRLI. Thus the host never established nvme connectivity
with the lpfc target.
By changing the rejection values (to Logical Busy/nothing more), the
initiator accepted the response and would retry the PRLI, resulting in
nvme connectivity.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When enabled for both SCSI and NVME support, and connected pt2pt to a
SCSI only target, the driver nodelist entry for the remote port is left
in PRLI_ISSUE state and no SCSI LUNs are discovered. Works fine if only
configured for SCSI support.
Error was due to some of the prli points still reflecting the need to
send only 1 PRLI. On a lot of fabric configs, targets were NVME only,
which meant the fabric-reported protocol attributes were only telling
the driver one protocol or the other. Thus things worked fine. With
pt2pt, the driver must send a PRLI for both protocols as there are no
hints on what the target supports. Thus pt2pt targets were hitting the
multiple PRLI issues.
Complete the dual PRLI support. Track explicitly whether scsi (fcp) or
nvme prli's have been sent. Accurately track protocol support detected
on each node as reported by the fabric or probed by PRLI traffic.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fix child-node lookup during probe, which ended up searching the whole
device tree depth-first starting at the parent rather than just matching
on its children.
Note that the original premature free of the parent node has already
been fixed separately.
Also note that this pattern of looking up the first child node with a
given property is rare enough that a generic helper is probably not
warranted.
Fixes: c97c4090ff72 ("soc: qcom: smsm: Add driver for Qualcomm SMSM") Fixes: 3e8b55411468 ("soc: qcom: smsm: fix of_node refcnting problem") Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Rob Clark <robdclark@gmail.com> Signed-off-by: Johan Hovold <johan@kernel.org> Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org> Signed-off-by: Andy Gross <andy.gross@linaro.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If md is NULL, tun_dst must be freed, otherwise it will cause memory
leak.
Fixes: 1a66a836da6 ("gre: add collect_md mode to ERSPAN tunnel") Cc: William Tu <u9012063@gmail.com> Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When erspan_rcv call return PACKET_REJECT, we shoudn't call ipgre_rcv to
process packets again, instead send icmp unreachable message in error
path.
Fixes: 84e54fe0a5ea ("gre: introduce native tunnel support for ERSPAN") Acked-by: William Tu <u9012063@gmail.com> Cc: William Tu <u9012063@gmail.com> Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LTP/udp6_ipsec_vti tests fail when sending large UDP datagrams over
ip6_vti that require fragmentation and the underlying device has an
MTU smaller than 1500 plus some extra space for headers. This happens
because ip6_vti, by default, sets MTU to ETH_DATA_LEN and not updating
it depending on a destination address or link parameter. Further
attempts to send UDP packets may succeed because pmtu gets updated on
ICMPV6_PKT_TOOBIG in vti6_err().
In case the lower device has larger MTU size, e.g. 9000, ip6_vti works
but not using the possible maximum size, output packets have 1500 limit.
The above cases require manual MTU setup after ip6_vti creation. However
ip_vti already updates MTU based on lower device with ip_tunnel_bind_dev().
Here is the example when the lower device MTU is set to 9000:
# ip a sh ltp_ns_veth2
ltp_ns_veth2@if7: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 ...
inet 10.0.0.2/24 scope global ltp_ns_veth2
inet6 fd00::2/64 scope global
# ip li add vti6 type vti6 local fd00::2 remote fd00::1
# ip li show vti6
vti6@NONE: <POINTOPOINT,NOARP> mtu 1500 ...
link/tunnel6 fd00::2 peer fd00::1
After the patch:
# ip li add vti6 type vti6 local fd00::2 remote fd00::1
# ip li show vti6
vti6@NONE: <POINTOPOINT,NOARP> mtu 8832 ...
link/tunnel6 fd00::2 peer fd00::1
Reported-by: Petr Vorel <pvorel@suse.cz> Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It is unlikely request_threaded_irq will fail, but if it does for some
reason we should clear iommu->pr_irq in the error path. Also
intel_svm_finish_prq shouldn't try to clean up the page request
interrupt if pr_irq is 0. Without these, if request_threaded_irq were
to fail the following occurs:
We generally leave the GPIO clock disabled, unless an interrupt is
requested or we're accessing IO registers. We forgot to do this for the
->get_direction() callback, which means we can sometimes [1] get
incorrect results [2] from, e.g., /sys/kernel/debug/gpio.
Enable the clock, so we get the right results!
[1] Sometimes, because many systems have 1 or mor interrupt requested on
each GPIO bank, so they always leave their clock on.
[2] Incorrect, meaning the register returns 0, and so we interpret that
as "input".
In case a platform only defaults a "default" set of pins, but not a
"sleep" set of pins, and this particular platform suspends and resumes
in a way that the pin states are not preserved by the hardware, when we
resume, we would call pinctrl_single_resume() -> pinctrl_force_default()
-> pinctrl_select_state() and the first thing we do is check that the
pins state is the same as before, and do nothing.
In order to fix this, decouple the actual state change from
pinctrl_select_state() and move it pinctrl_commit_state(), while keeping
the p->state == state check in pinctrl_select_state() not to change the
caller assumptions. pinctrl_force_sleep() and pinctrl_force_default()
are updated to bypass the state check by calling pinctrl_commit_state().
[Linus Walleij]
The forced pin control states are currently only used in some pin
controller drivers that grab their own reference to their own pins.
This is equal to the pin control hogs: pins taken by pin control
devices since there are no corresponding device in the Linux device
hierarchy, such as memory controller lines or unused GPIO lines,
or GPIO lines that are used orthogonally from the GPIO subsystem
but pincontrol-wise managed as hogs (non-strict mode, allowing
simultaneous use by GPIO and pin control). For this case forcing
the state from the drivers' suspend()/resume() callbacks makes
sense and should semantically match the name of the function.
Fixes: 6e5e959dde0d ("pinctrl: API changes to support multiple states per device") Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com> Signed-off-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Two orthogonal changesets caused a breakage at a printk
inside davinci. Commit a2d17962c9ca
("[media] davinci: Switch from V4L2 OF to V4L2 fwnode")
made davinci to use struct fwnode_handle instead of
struct device_node. Commit 68d9c47b1679
("media: Convert to using %pOF instead of full_name")
changed the printk to not use ->full_name, but, instead,
to rely on %pOF.
With both patches applied, the Kernel will do the wrong
thing, as warned by smatch:
drivers/media/platform/davinci/vpif_capture.c:1399 vpif_async_bound() error: '%pOF' expects argument of type 'struct device_node*', argument 5 has type 'void*'
So, change the logic to actually print the device name
that was obtained before the print logic.
Fixes: 68d9c47b1679 ("media: Convert to using %pOF instead of full_name") Fixes: a2d17962c9ca ("[media] davinci: Switch from V4L2 OF to V4L2 fwnode") Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com> Acked-by: Lad, Prabhakar <prabhakar.csengg@gmail.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
->get_msi() now checks MSI_EN bit in the MSI CAPABILITY register to
find whether the host supports MSI instead of using the
MSI ADDRESS in the MSI CAPABILITY register.
This fixes the issue with the following sequence
'modprobe pci_endpoint_test' enables MSI
'rmmod pci_endpoint_test' disables MSI but MSI address (in EP's
capability register) has a valid value
'modprobe pci_endpoint_test no_msi=1' - Since MSI address (in EP's
capability register) has a valid value (set during the previous
insertion of the module), EP thinks host supports MSI.
Fixes: f8aed6ec624f ("PCI: dwc: designware: Add EP mode support") Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The CoreSight TPIU should be disabled when tracing to other sinks to allow
them to operate at full bandwidth.
This patch fixes tpiu_disable_hw() to correctly disable the TPIU by
configuring the TPIU to stop on flush, initiating a manual flush, waiting
for the flush to complete and then waits for the TPIU to indicate it has
stopped.
In case that CONFIG_SLUB_DEBUG is on and pty is used, races between
release_one_tty and flush_to_ldisc work threads may happen and lead
to use-after-free condition on tty->link->port. Because SLUB_DEBUG
is turned on, freed tty->link->port is filled with POISON_FREE value.
So far without SLUB_DEBUG, port was filled with zero and flush_to_ldisc
could return without a problem by checking if tty is NULL.
CPU 0 CPU 1
----- -----
release_tty pty_write
cancel_work_sync(tty) to = tty->link
tty_kref_put(tty->link) tty_schedule_flip(to->port)
<< workqueue >> ...
release_one_tty ...
pty_cleanup ...
kfree(tty->link->port) << workqueue >>
flush_to_ldisc
tty = READ_ONCE(port->itty)
tty is 0x6b6b6b6b6b6b6b6b
!!PANIC!! access tty->ldisc
Check the status of the DMM engine after it is reported that the
transaction was completed as in rare cases the engine might not reached a
working state.
The wait_status() will print information in case the DMM is not reached the
expected state and the dmm_txn_commit() will return with an error code to
make sure that we are not continuing with a broken setup.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to SD spec 3.00 3.6.1 signal voltage switch
procedure step 6~8,
(6) Set 1.8V Signal Enable in the Host Control 2 register.
(7) Wait 5ms. 1.8V voltage regulator shall be stable within this period.
(8) If 1.8V Signal Enable is cleared by Host Controller, go to step (12).
Host should wait 5ms after set 1.8V signal enable bit in
Host Control 2 register and check if 1.8V is stable or not.
But current code checks this bit right after set it.
On some platforms with xenon controller found the bit is
cleared right away and host reports "1.8V regulator output
did not became stable" and 5ms delay can help.
Implement voltage_switch callback for xenon controller to add 5ms
delay to make sure the 1.8V signal enable bit is set by controller.
Set the resource type when we reserve VGA-related I/O port resources.
The resource code doesn't actually look at the type, so it inserts
resources without a type in the tree correctly even without this change.
But if we ever print a resource without a type, it looks like this:
vga+ [??? 0x000003c0-0x000003df flags 0x0]
Setting the type means it will be printed correctly as:
isert commands that failed during isert_rdma_rw_ctx_post() are queued to
Queue-Full(QF) queue and are scheduled to be reposted during queue-full
queue processing. During this reposting, the rdma contexts are initialised
again in isert_rdma_rw_ctx_post(), which is leaking significant memory.
The ipoib path database is organized around DGIDs from the LLADDR, but the
SA is free to return a different GID when asked for path. This causes a
bug because the SA's modified DGID is copied into the database key, even
though it is no longer the correct lookup key, causing a memory leak and
other malfunctions.
Ensure the database key does not change after the SA query completes.
Demonstration of the bug is as follows
ipoib wants to send to GID fe80:0000:0000:0000:0002:c903:00ef:5ee2, it
creates new record in the DB with that gid as a key, and issues a new
request to the SM.
Now, the SM from some reason returns path-record with other SGID (for
example, 2001:0000:0000:0000:0002:c903:00ef:5ee2 that contains the local
subnet prefix) now ipoib will overwrite the current entry with the new
one, and if new request to the original GID arrives ipoib will not find
it in the DB (was overwritten) and will create new record that in its
turn will also be overwritten by the response from the SM, and so on
till the driver eats all the device memory.
Signed-off-by: Erez Shitrit <erezsh@mellanox.com> Signed-off-by: Leon Romanovsky <leon@kernel.org> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Driver calls request_firmware() whenever the device is opened for the
first time. As the device gets opened and closed, dev->num_inst == 1
is true several times. This is not necessary since the firmware is saved
in the fw_buf. s5p_mfc_load_firmware() copies the buffer returned by
the request_firmware() to dev->fw_buf.
fw_buf sticks around until it gets released from s5p_mfc_remove(), hence
there is no need to keep requesting firmware and copying it to fw_buf.
This might have been overlooked when changes are made to free fw_buf from
the device release interface s5p_mfc_release().
Fix s5p_mfc_load_firmware() to call request_firmware() once and keep state.
Change _probe() to load firmware once fw_buf has been allocated.
s5p_mfc_open() and it continues to call s5p_mfc_load_firmware() and init
hardware which is the step where firmware is written to the device.
This addresses the mfc_mutex contention due to repeated request_firmware()
calls from open() in the following circular locking warning:
[ 552.194115] qtdemux0:sink/2710 is trying to acquire lock:
[ 552.199488] (&dev->mfc_mutex){+.+.}, at: [<bf145544>] s5p_mfc_mmap+0x28/0xd4 [s5p_mfc]
[ 552.207459]
but task is already holding lock:
[ 552.213264] (&mm->mmap_sem){++++}, at: [<c01df2e4>] vm_mmap_pgoff+0x44/0xb8
[ 552.220284]
which lock already depends on the new lock.
The detection of a PHY changed in commit e98a3aabf85f ("mdio_bus: don't
return NULL from mdiobus_scan()") which now causes sfp to print an
error message. Update for this change.
Fixes: 73970055450e ("sfp: add SFP module support") Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The EEPROM reading was trying to read from the second EEPROM address
if we requested the last byte from the SFF8079 EEPROM, which caused a
failure when the second EEPROM is not present. Discovered with a
S-RJ01 SFP module. Fix this.
Fixes: 73970055450e ("sfp: add SFP module support") Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit df04abfd181a ("fs/proc/kcore.c: Add bounce buffer for ktext data")
this adds a bounce buffer when reading memory via /dev/mem. This
is needed to allow kernel text memory to be read out when built with
CONFIG_HARDENED_USERCOPY (which refuses to read out kernel text) and
without CONFIG_STRICT_DEVMEM (which would have refused to read any RAM
contents at all).
Since this build configuration isn't common (most systems with
CONFIG_HARDENED_USERCOPY also have CONFIG_STRICT_DEVMEM), this also tries
to inform Kconfig about the recommended settings.
This patch is modified from Brad Spengler/PaX Team's changes to /dev/mem
code in the last public patch of grsecurity/PaX based on my understanding
of the code. Changes or omissions from the original code are mine and
don't reflect the original grsecurity/PaX code.
The Weibu F3C MiniPC has an onboard AP6255 module, presenting
two SDIO functions on a single MMC host (Bluetooth/btsdio and
WiFi/brcmfmac), and the mmc layer correctly detects this as
non-removable.
After suspend/resume, the wifi and bluetooth interfaces disappear
and do not get probed again.
The conditions here are:
1. During suspend, we reach mmc_pm_notify()
2. mmc_pm_notify() calls mmc_sdio_pre_suspend() to see if we can
suspend the SDIO host. However, mmc_sdio_pre_suspend() returns
-ENOSYS because btsdio_driver does not have a suspend method.
3. mmc_pm_notify() proceeds to remove the card
4. Upon resume, mmc_rescan() does nothing with this host, because of
the rescan_entered check which aims to only scan a non-removable
device a single time (i.e. during boot).
Fix the loss of functionality by detecting that we are unable to
suspend a non-removable host, so avoid the forced removal in that
case. The comment above this function already indicates that this
code was only intended for removable devices.
Add a check to ensure iowrite64 is only used if it is atomic.
It was decided in [1] that the tilcdc driver should not be using an
atomic operation (so it was left out of this patchset). However, it turns
out that through the drm code, a nonatomic header is actually included:
include/linux/io-64-nonatomic-lo-hi.h
is included from include/drm/drm_os_linux.h:9:0,
from include/drm/drmP.h:74,
from include/drm/drm_modeset_helper.h:26,
from include/drm/drm_atomic_helper.h:33,
from drivers/gpu/drm/tilcdc/tilcdc_crtc.c:19:
And thus, without this change, this patchset would inadvertantly
change the behaviour of the tilcdc driver.
pkt_xfer should be used for protocol v3, and cmd_xfer otherwise. We had
one instance of these functions correct, but not the second, fall-back
case. We use the fall-back only when the first command returns an
IN_PROGRESS status, which is only used on some EC firmwares where we
don't want to constantly poll the bus, but instead back off and
sleep/retry for a little while.
If handle_boot_enabled is set to 0, the watchdog driver module use
counter will not be increased and kref_get() will not be called when
registering the watchdog. Subsequently, on open, this does not happen
either because the code believes that it was already done because the
hardware watchdog is marked as running.
We could introduce a state variable to indicate this state, but let's
just increase the module use counter and call kref_get() unconditionally
if the hardware watchdog is running when a driver is registering itself
to keep the code simple.
Fixes: 2501b015313fe ("watchdog: core: add option to avoid early ...") Cc: Sebastian Reichel <sebastian.reichel@collabora.co.uk> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Wim Van Sebroeck <wim@iguana.be> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a watchdog driver's open function sets WDOG_HW_RUNNING with the
expectation that the watchdog can not be stopped, but then stops the
watchdog anyway in its stop function, kref_get() wil not be called in
watchdog_open(). If the watchdog then stops on close, WDOG_HW_RUNNING
will be cleared and kref_put() will be called, causing a kref imbalance.
As result the character device data structure will be released, which in
turn will cause the system to crash on the next call to watchdog_open().
drivers/platform/chrome/cros_ec_sysfs.c: In function 'show_ec_version':
drivers/platform/chrome/cros_ec_sysfs.c:190:12: error: array subscript 4294967295 is above array bounds of 'uint8_t[]' [-Werror=array-bounds]
This changes the code back to what it did before changing to a
zero-length array structure.
Fixes: a841178445bb ("mfd: cros_ec: Use a zero-length array for command data") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Benson Leung <bleung@chromium.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If there is no clock rate for uartclk defined, disable the previously
enabled clock again.
Found by Linux Driver Verification project (linuxtesting.org).
Fixes: 23f5b3fdd04e serial: 8250_dw: only get the clock rate in one place Signed-off-by: Stefan Potyra <Stefan.Potyra@elektrobit.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 3840ed9548f7 ("tty: goldfish: Implement support for kernel
'earlycon' parameter") breaks an allmodconfig config on x86:
| LD vmlinux.o
| MODPOST vmlinux.o
|drivers/tty/serial/earlycon.o: In function `parse_options':
|drivers/tty/serial/earlycon.c:97: undefined reference to `uart_parse_earlycon'
|Makefile:1005: recipe for target 'vmlinux' failed
earlycon.c::parse_options() invokes uart_parse_earlycon() from serial_core.c
which is compiled=m because GOLDFISH_TTY itself (and most others) are =m.
To avoid that, I'm adding the _CONSOLE config option which is selected if the
GOLDFISH module itself is =y since it doesn't need the early bits for the =m
case (other drivers do the same dance).
The alternative would be to move uart_parse_earlycon() from
serial_core.c to earlycon.c (we don't have that many users of that
function).
Fixes: 3840ed9548f7 ("tty: goldfish: Implement support for kernel
'earlycon' parameter") Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Miodrag Dinic <miodrag.dinic@mips.com> Acked-by: Miodrag Dinic <miodrag.dinic@mips.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It has been reported that the dummy byte we add to avoid
ZLPs can be forwarded by the modem to the PGW/GGSN, and that
some operators will drop the connection if this happens.
In theory, QMI devices are based on CDC ECM and should as such
both support ZLPs and silently ignore the dummy byte. The latter
assumption failed. Let's test out the first.
Signed-off-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On faster CPUs a delay is required after the resume command and the restart command. Without the delay, the restart command often returns -EREMOTEIO and the Si2168 does not restart.
Note that this patch fixes the same issue as https://patchwork.linuxtv.org/patch/44304/, but I believe my udelay() fix addresses the actual problem.
Data packets are not sent by STA in case of STA joined to
non QOS AP (WMM disabled AP). This is happening because of STA
is sending data packets to firmware from host with qos enabled
along with non qos queue value(TID = 16).
Due to qos enabled, firmware is discarding the packet.
This patch fixes this issue by updating the qos based on station
WME capability field if WMM is disabled in AP.
This patch is required by 10.4 family chipsets like
QCA4019/QCA9888/QCA9884/QCA99X0.
Firmware Versoin : 10.4-3.5.1-00018.
For 10.2.4 family chipsets QCA988X/QCA9887 and QCA6174 this patch
has no effect.
Signed-off-by: Balaji Pothunoori <bpothuno@qti.qualcomm.com> Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In rtl_rx_ampdu_apply(), when rtlpriv->cfg->ops->get_btc_status()
returns false, RT_TRACE() is called with the values of variables
reject_agg and agg_size, which have not been initialized.
Always initialize these variables in order to prevent using
uninitialized values.
This issue has been found with clang. The compiler reported:
drivers/net/wireless/realtek/rtlwifi/base.c:1665:6: error: variable
'agg_size' is used uninitialized whenever 'if' condition is false
[-Werror,-Wsometimes-uninitialized]
if (rtlpriv->cfg->ops->get_btc_status())
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/net/wireless/realtek/rtlwifi/base.c:1671:31: note:
uninitialized use occurs here
reject_agg, ctrl_agg_size, agg_size);
^~~~~~~~
drivers/net/wireless/realtek/rtlwifi/base.c:1665:6: error: variable
'reject_agg' is used uninitialized whenever 'if' condition
is false [-Werror,-Wsometimes-uninitialized]
if (rtlpriv->cfg->ops->get_btc_status())
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/net/wireless/realtek/rtlwifi/base.c:1671:4: note:
uninitialized use occurs here
reject_agg, ctrl_agg_size, agg_size);
^~~~~~~~~~
Fixes: 2635664e6e4a ("rtlwifi: Add rx ampdu cfg for btcoexist.") Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reset the driver current tx read/write index to zero when inactiveps
nic out of sync with HW state. Wrong driver tx read/write index will
cause Tx fail.
Signed-off-by: Tsang-Shian Lin <thlin@realtek.com> Signed-off-by: Ping-Ke Shih <pkshih@realtek.com> Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net> Cc: Yan-Hsuan Chuang <yhchuang@realtek.com> Cc: Birming Chiu <birming@realtek.com> Cc: Shaofu <shaofu@realtek.com> Cc: Steven Ting <steventing@realtek.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The spi_master.setup() callback must not change configuration registers,
as that could corrupt I/O that is in progress for other SPI slaves.
The only exception is the configuration of the native chip select
polarity in SPI master mode, as a wrong chip select polarity will cause
havoc during all future transfers to any other SPI slave.
Hence stop writing to registers in sh_msiof_spi_setup(), unless it is
the first call for a controller using a native chip select, or unless
native chip select polarity has changed (note that you'll loose anyway
if I/O is in progress). Even then, only do what is strictly necessary,
instead of calling sh_msiof_spi_set_pin_regs().
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The max should be 31 MB on host with NVSP version > 2.
On legacy hosts (NVSP version <=2) only 15 MB receive buffer is allowed,
otherwise the buffer request will be rejected by the host, resulting
vNIC not coming up.
The NVSP version is only available after negotiation. So, we add the
limit checking for legacy hosts in netvsc_init_buf().
Fixes: 5023a6db73196 ("netvsc: increase default receive buffer size") Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/infiniband/core/iwpm_util.c: In function ‘iwpm_send_mapinfo’:
drivers/infiniband/core/iwpm_util.c:647: warning: ‘ret’ may be used uninitialized in this function
Indeed, if nl_client is not found in any of the scanned has buckets, ret
will be used uninitialized.
Preinitialize ret to -EINVAL to fix this.
Fixes: 30dc5e63d6a5ad24 ("RDMA/core: Add support for iWARP Port Mapper user space service") Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Reviewed-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
get_pages doesn't keep a reference of the pages allocated
when it fails later in the code path. This can lead to
a memory leak. Keep reference of the allocated pages so
that it can be freed when msm_gem_free_object gets called
later during cleanup.
Signed-off-by: Prakash Kamliya <pkamliya@codeaurora.org> Signed-off-by: Sharat Masetty <smasetty@codeaurora.org> Signed-off-by: Rob Clark <robdclark@gmail.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
_channel_ is being dereferenced before it is null checked, hence there is a
potential null pointer dereference. Fix this by moving the pointer dereference
after _channel_ has been null checked.
This issue was detected with the help of Coccinelle.
The commit e948bc8fbee0 ("cpufreq: Cap the default transition delay
value to 10 ms") caused a regression on EPIA-M min-ITX computer where
shutdown or reboot hangs occasionally with a print message like:
longhaul: Warning: Timeout while waiting for idle PCI bus
cpufreq: __target_index: Failed to change cpu frequency: -16
This probably happens because the cpufreq governor tries to change the
frequency of the CPU faster than allowed by the hardware.
Before the above commit, the default transition delay was set to 200 ms
for a transition_latency of 200000 ns. Lets revert back to that
transition delay value to fix it. Note that several other transition
delay values were tested like 20 ms and 30 ms and none of them have
resolved system hang issue completely.
Fixes: e948bc8fbee0 (cpufreq: Cap the default transition delay value to 10 ms) Reported-by: Meelis Roos <mroos@linux.ee> Suggested-by: Rafael J. Wysocki <rjw@rjwysocki.net> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <alexander.levin@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Assuming that the original code idea was to enable in-band sleeping
only if the setup_rome method returns succes and run in 'standard'
mode otherwise, we should not return setup_rome return value which
makes qca_setup fail if no rampatch/nvm file found.
This fixes BT issue on the dragonboard-820C p4 which includes the
following QCA controller:
hci0: Product:0x00000008
hci0: Patch :0x00000111
hci0: ROM :0x00000302
hci0: SOC :0x00000044
Since there is no rampatch for this controller revision, just make
it work as is.
There is a lock odering problem between mmap_sem and ashmem_mutex causing
a lockdep splat[1] during a syzcaller test. This patch fixes the problem
by move copy_from_user out of ashmem_mutex.
Because of the shifting around of code in qla2x00_probe_one recently,
failures during adapter initialization can lead to problems, i.e. NULL
pointer crashes and doubly freed data structures which cause eventual
panics.
This V2 version makes the relevant memory free routines idempotent, so
repeat calls won't cause any harm. I also removed the problematic
probe_init_failed exit point as it is not needed.
Fixes: d64d6c5671db ("scsi: qla2xxx: Fix NULL pointer crash due to probe failure") Signed-off-by: Bill Kuzeja <william.kuzeja@stratus.com> Acked-by: Himanshu Madhani <himanshu.madhani@cavium.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 3515832cc614 ("scsi: qla2xxx: Reset the logo flag, after target
re-login.")fixed the target re-login after session relogin is complete,
but missed out the qlt_free_session_done() path.
This patch clears send_els_logo flag in qlt_free_session_done()
callback.
[mkp: checkpatch]
Fixes: 3515832cc614 ("scsi: qla2xxx: Reset the logo flag, after target re-login.") Signed-off-by: Himanshu Madhani <hmadhani@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
when processing iocb in a timeout case, driver was trying to log messages
without verifying if the fcport structure could have valid data. This
results in a NULL pointer access.
Fixes: 726b85487067("qla2xxx: Add framework for async fabric discovery") Signed-off-by: Quinn Tran <quinn.tran@cavium.com> Signed-off-by: Himanshu Madhani <himanshu.madhani@cavium.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit addc3fa74e5b ("Btrfs: Fix the problem that the dirty flag of dev
stats is cleared") reworked the way device stats changes are tracked. A
new atomic dev_stats_ccnt counter was introduced which is incremented
every time any of the device stats counters are changed. This serves as
a flag whether there are any pending stats changes. However, this patch
only partially implemented the correct memory barriers necessary:
- It only ordered the stores to the counters but not the reads e.g.
btrfs_run_dev_stats
- It completely omitted any comments documenting the intended design and
how the memory barriers pair with each-other
This patch provides the necessary comments as well as adds a missing
smp_rmb in btrfs_run_dev_stats. Furthermore since dev_stats_cnt is only
a snapshot at best there was no point in reading the counter twice -
once in btrfs_dev_stats_dirty and then again when assigning stats_cnt.
Just collapse both reads into 1.
Fixes: addc3fa74e5b ("Btrfs: Fix the problem that the dirty flag of dev stats is cleared") Signed-off-by: Nikolay Borisov <nborisov@suse.com> Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
WARNING: CPU: 3 PID: 18172 at fs/btrfs/backref.c:1391 find_parent_nodes+0xc41/0x14e0
Modules linked in: [...]
CPU: 3 PID: 18172 Comm: bees Tainted: G D W L 4.11.9-zb64+ #1
Hardware name: System manufacturer System Product Name/M5A78L-M/USB3, BIOS 2101 12/02/2014
Call Trace:
dump_stack+0x85/0xc2
__warn+0xd1/0xf0
warn_slowpath_null+0x1d/0x20
find_parent_nodes+0xc41/0x14e0
__btrfs_find_all_roots+0xad/0x120
? extent_same_check_offsets+0x70/0x70
iterate_extent_inodes+0x168/0x300
iterate_inodes_from_logical+0x87/0xb0
? iterate_inodes_from_logical+0x87/0xb0
? extent_same_check_offsets+0x70/0x70
btrfs_ioctl+0x8ac/0x2820
? lock_acquire+0xc2/0x200
do_vfs_ioctl+0x91/0x700
? __fget+0x112/0x200
SyS_ioctl+0x79/0x90
entry_SYSCALL_64_fastpath+0x23/0xc6
? trace_hardirqs_off_caller+0x1f/0x140
Starting with v4.14 (specifically 86d5f9944252 ("btrfs: convert prelimary
reference tracking to use rbtrees")) the WARN_ON occurs three orders of
magnitude more frequently--almost once per second while running workloads
like bees.
Replace the WARN_ON() with a comment rationale for its removal.
The rationale is paraphrased from an explanation by Edmund Nadolski
<enadolski@suse.de> on the linux-btrfs mailing list.
Commit 4fde46f0cc71 ("Btrfs: free the stale device") introduced
btrfs_free_stale_device which iterates the device lists for all
registered btrfs filesystems and deletes those devices which aren't
mounted. In a btrfs_devices structure has only 1 device attached to it
and it is unused then btrfs_free_stale_devices will proceed to also free
the btrfs_fs_devices struct itself. Currently this leads to a use after
free since list_for_each_entry will try to perform a check on the
already freed memory to see if it has to terminate the loop.
The fix is to use 'break' when we know we are freeing the current
fs_devs.
Fixes: 4fde46f0cc71 ("Btrfs: free the stale device") Signed-off-by: Nikolay Borisov <nborisov@suse.com> Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In case of using DUP, we search for enough unallocated disk space on a
device to hold two stripes.
The devices_info[ndevs-1].max_avail that holds the amount of unallocated
space found is directly assigned to stripe_size, while it's actually
twice the stripe size.
Later on in the code, an unconditional division of stripe_size by
dev_stripes corrects the value, but in the meantime there's a check to
see if the stripe_size does not exceed max_chunk_size. Since during this
check stripe_size is twice the amount as intended, the check will reduce
the stripe_size to max_chunk_size if the actual correct to be used
stripe_size is more than half the amount of max_chunk_size.
The unconditional division later tries to correct stripe_size, but will
actually make sure we can't allocate more than half the max_chunk_size.
Fix this by moving the division by dev_stripes before the max chunk size
check, so it always contains the right value, instead of putting a duct
tape division in further on to get it fixed again.
Since in all other cases than DUP, dev_stripes is 1, this change only
affects DUP.
Other attempts in the past were made to fix this:
* 37db63a400 "Btrfs: fix max chunk size check in chunk allocator" tried
to fix the same problem, but still resulted in part of the code acting
on a wrongly doubled stripe_size value.
* 86db25785a "Btrfs: fix max chunk size on raid5/6" unintentionally
broke this fix again.
The real problem was already introduced with the rest of the code in 73c5de0051.
The user visible result however will be that the max chunk size for DUP
will suddenly double, while it's actually acting according to the limits
in the code again like it was 5 years ago.
Reported-by: Naohiro Aota <naohiro.aota@wdc.com> Link: https://www.spinics.net/lists/linux-btrfs/msg69752.html Fixes: 73c5de0051 ("btrfs: quasi-round-robin for chunk allocation") Fixes: 86db25785a ("Btrfs: fix max chunk size on raid5/6") Signed-off-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com> Reviewed-by: David Sterba <dsterba@suse.com>
[ update comment ] Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
wrote 65536/65536 bytes at offset 0
64 KiB, 4 ops; 0.0000 sec (121.359 MiB/sec and 7766.9903 ops/sec)
/media/scratch/file5:
EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS
0: [0..127]: 24576..24703 128 0x2001
/media/scratch/file5:
EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS
0: [0..127]: 24576..24703 128 0x1
This is because btrfs_check_shared calls find_parent_nodes
repeatedly in a loop, passing a share_check struct to report
the count of shared extent. But btrfs_check_shared does not
re-initialize the count value to zero for subsequent calls
from the loop, resulting in a false share count value. This
is a regressive behavior from 4.13.
With proper re-initialization the test result is as follows:
wrote 65536/65536 bytes at offset 0
64 KiB, 4 ops; 0.0000 sec (110.035 MiB/sec and 7042.2535 ops/sec)
/media/scratch/file5:
EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS
0: [0..127]: 24576..24703 128 0x1
/media/scratch/file5:
EXT: FILE-OFFSET BLOCK-RANGE TOTAL FLAGS
0: [0..127]: 24576..24703 128 0x1
which corrects the regression.
Fixes: 3ec4d3238ab ("btrfs: allow backref search checks for shared extents") Signed-off-by: Edmund Nadolski <enadolski@suse.com>
[ add text from cover letter to changelog ] Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This is reproducible in a cycle, where a series of writes is followed by
SCSI device delete command. The test may take up to few minutes.
Fixes: 74d46992e0d9 ("block: replace bi_bdev with a gendisk pointer and partitions index")
[ no signed-off-by provided ]
Author: Dmitriy Gorokh <Dmitriy.Gorokh@wdc.com> Reviewed-by: Liu Bo <bo.li.liu@oracle.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When struct its_device instances are created, the nr_ites member
will be set to a power of 2 that equals or exceeds the requested
number of MSIs passed to the msi_prepare() callback. At the same
time, the LPI map is allocated to be some multiple of 32 in size,
where the allocated size may be less than the requested size
depending on whether a contiguous range of sufficient size is
available in the global LPI bitmap.
This may result in the situation where the nr_ites < nr_lpis, and
since nr_ites is what we program into the hardware when we map the
device, the additional LPIs will be non-functional.
For bog standard hardware, this does not really matter. However,
in cases where ITS device IDs are shared between different PCIe
devices, we may end up allocating these additional LPIs without
taking into account that they don't actually work.
So let's make nr_ites at least 32. This ensures that all allocated
LPIs are 'live', and that its_alloc_device_irq() will fail when
attempts are made to allocate MSIs beyond what was allocated in
the first place.
rvt_mregion uses percpu_ref for reference counting and RCU to protect
accesses from lkey_table. When a rvt_mregion needs to be freed, it
first gets unregistered from lkey_table and then rvt_check_refs() is
called to wait for in-flight usages before the rvt_mregion is freed.
rvt_check_refs() seems to have a couple issues.
* It has a fast exit path which tests percpu_ref_is_zero(). However,
a percpu_ref reading zero doesn't mean that the object can be
released. In fact, the ->release() callback might not even have
started executing yet. Proceeding with freeing can lead to
use-after-free.
* lkey_table is RCU protected but there is no RCU grace period in the
free path. percpu_ref uses RCU internally but it's sched-RCU whose
grace periods are different from regular RCU. Also, it generally
isn't a good idea to depend on internal behaviors like this.
To address the above issues, this patch removes the fast exit and adds
an explicit synchronize_rcu().
While converting ioctx index from a list to a table, db446a08c23d
("aio: convert the ioctx list to table lookup v3") missed tagging
kioctx_table->table[] as an array of RCU pointers and using the
appropriate RCU accessors. This introduces a small window in the
lookup path where init and access may race.
Mark kioctx_table->table[] with __rcu and use the approriate RCU
accessors when using the field.
Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: Jann Horn <jannh@google.com> Fixes: db446a08c23d ("aio: convert the ioctx list to table lookup v3") Cc: Benjamin LaHaise <bcrl@kvack.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: stable@vger.kernel.org # v3.12+ Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While fixing refcounting, e34ecee2ae79 ("aio: Fix a trinity splat")
incorrectly removed explicit RCU grace period before freeing kioctx.
The intention seems to be depending on the internal RCU grace periods
of percpu_ref; however, percpu_ref uses a different flavor of RCU,
sched-RCU. This can lead to kioctx being freed while RCU read
protected dereferences are still in progress.
Fix it by updating free_ioctx() to go through call_rcu() explicitly.
In case when dentry passed to lock_parent() is protected from freeing only
by the fact that it's on a shrink list and trylock of parent fails, we
could get hit by __dentry_kill() (and subsequent dentry_kill(parent))
between unlocking dentry and locking presumed parent. We need to recheck
that dentry is alive once we lock both it and parent *and* postpone
rcu_read_unlock() until after that point. Otherwise we could return
a pointer to struct dentry that already is rcu-scheduled for freeing, with
->d_lock held on it; caller's subsequent attempt to unlock it can end
up with memory corruption.
The vgic code is trying to be clever when injecting GICv2 SGIs,
and will happily populate LRs with the same interrupt number if
they come from multiple vcpus (after all, they are distinct
interrupt sources).
Unfortunately, this is against the letter of the architecture,
and the GICv2 architecture spec says "Each valid interrupt stored
in the List registers must have a unique VirtualID for that
virtual CPU interface.". GICv3 has similar (although slightly
ambiguous) restrictions.
This results in guests locking up when using GICv2-on-GICv3, for
example. The obvious fix is to stop trying so hard, and inject
a single vcpu per SGI per guest entry. After all, pending SGIs
with multiple source vcpus are pretty rare, and are mostly seen
in scenario where the physical CPUs are severely overcomitted.
But as we now only inject a single instance of a multi-source SGI per
vcpu entry, we may delay those interrupts for longer than strictly
necessary, and run the risk of injecting lower priority interrupts
in the meantime.
In order to address this, we adopt a three stage strategy:
- If we encounter a multi-source SGI in the AP list while computing
its depth, we force the list to be sorted
- When populating the LRs, we prevent the injection of any interrupt
of lower priority than that of the first multi-source SGI we've
injected.
- Finally, the injection of a multi-source SGI triggers the request
of a maintenance interrupt when there will be no pending interrupt
in the LRs (HCR_NPIE).
At the point where the last pending interrupt in the LRs switches
from Pending to Active, the maintenance interrupt will be delivered,
allowing us to add the remaining SGIs using the same process.
On guest exit, and when using GICv2 on GICv3, we use a dsb(st) to
force synchronization between the memory-mapped guest view and
the system-register view that the hypervisor uses.
This is incorrect, as the spec calls out the need for "a DSB whose
required access type is both loads and stores with any Shareability
attribute", while we're only synchronizing stores.
We also lack an isb after the dsb to ensure that the latter has
actually been executed before we start reading stuff from the sysregs.
The fix is pretty easy: turn dsb(st) into dsb(sy), and slap an isb()
just after.
Cc: stable@vger.kernel.org Fixes: f68d2b1b73cc ("arm64: KVM: Implement vgic-v3 save/restore") Acked-by: Christoffer Dall <cdall@kernel.org> Reviewed-by: Andre Przywara <andre.przywara@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The KVM IDMAP is a mapping of a statically allocated kernel structure,
and so printing its physical address leaks the physical placement of
the kernel when physical KASLR in effect. So change the kvm_info() to
kvm_debug() to remove it from the log output.
While at it, trim the output a bit more: IRQ numbers can be found in
/proc/interrupts, and the HYP VA and vgic-v2 lines are not highly
informational either.
Cc: <stable@vger.kernel.org> Acked-by: Will Deacon <will.deacon@arm.com> Acked-by: Christoffer Dall <cdall@kernel.org> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On nfsv2 and nfsv3 the nfs server can export subsets of the same
filesystem and report the same filesystem identifier, so that the nfs
client can know they are the same filesystem. The subsets can be from
disjoint directory trees. The nfsv2 and nfsv3 filesystems provides no
way to find the common root of all directory trees exported form the
server with the same filesystem identifier.
The practical result is that in struct super s_root for nfs s_root is
not necessarily the root of the filesystem. The nfs mount code sets
s_root to the root of the first subset of the nfs filesystem that the
kernel mounts.
This effects the dcache invalidation code in generic_shutdown_super
currently called shrunk_dcache_for_umount and that code for years
has gone through an additional list of dentries that might be dentry
trees that need to be freed to accomodate nfs.
When I wrote path_connected I did not realize nfs was so special, and
it's hueristic for avoiding calling is_subdir can fail.
The practical case where this fails is when there is a move of a
directory from the subtree exposed by one nfs mount to the subtree
exposed by another nfs mount. This move can happen either locally or
remotely. With the remote case requiring that the move directory be cached
before the move and that after the move someone walks the path
to where the move directory now exists and in so doing causes the
already cached directory to be moved in the dcache through the magic
of d_splice_alias.
If someone whose working directory is in the move directory or a
subdirectory and now starts calling .. from the initial mount of nfs
(where s_root == mnt_root), then path_connected as a heuristic will
not bother with the is_subdir check. As s_root really is not the root
of the nfs filesystem this heuristic is wrong, and the path may
actually not be connected and path_connected can fail.
The is_subdir function might be cheap enough that we can call it
unconditionally. Verifying that will take some benchmarking and
the result may not be the same on all kernels this fix needs
to be backported to. So I am avoiding that for now.
Filesystems with snapshots such as nilfs and btrfs do something
similar. But as the directory tree of the snapshots are disjoint
from one another and from the main directory tree rename won't move
things between them and this problem will not occur.
Cc: stable@vger.kernel.org Reported-by: Al Viro <viro@ZenIV.linux.org.uk> Fixes: 397d425dc26d ("vfs: Test for and handle paths that are unreachable from their mnt_root") Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>