As long as I know, in userspace, '%c' format on printing format for
tracepoint is replaced with '>c<' by existent tracing program; i.g.
'perf-trace' and 'trace-cmd'. This is inconvenient.
This commit replaces the format with '%s'. The length of letters in the
format string is not changed, thus this commit doesn't increase object
size.
In theory, I should work for improvements of these tracing programs, but
here I'd like to save my time to work for the other projects.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In design of ALSA PCM core, status and control data for runtime of ALSA
PCM substream are shared between kernel/user spaces by page frame
mapping with read-only attribute. Both of hardware-side and
application-side position on PCM buffer are maintained as a part of
the status data. In a view of ALSA PCM application, these two positions
can be updated by executing ioctl(2) with some commands.
There's an event of tracepoint for hardware-side position; 'hwptr'.
On the other hand, no events for application-side position. This commit
adds a new event for this purpose; 'applptr'. When the application-side
position is changed in kernel space, this event is probed with useful
information for developers.
I note that the event is not probed for all of ALSA PCM applications, When
applications are written by read/write programming scenario, the event is
surely probed. The applications execute ioctl(2) with
SNDRV_PCM_IOCTL_[READ|WRITE][N/I]_FRAMES to read/write any PCM frame, then
ALSA PCM core updates the application-side position in kernel land.
However, when applications are written by mmap programming scenario, if
maintaining the application side position in kernel space accurately,
applications should voluntarily execute ioctl(2) with
SNDRV_PCM_IOCTL_SYNC_PTR to commit the number of handled PCM frames. If
not voluntarily, the application-side position is not changed, thus the
added event is not probed.
There's a loophole, using architectures to which ALSA PCM core judges
non cache coherent. In this case, the status and control data is not mapped
into processe's VMA for any applications. Userland library, alsa-lib, is
programmed for this case. It executes ioctl(2) with
SNDRV_PCM_IOCTL_SYNC_PTR command every time to requiring the status and
control data.
ARM is such an architecture. Below is an example with serial sound interface
(ssi) on i.mx6 quad core SoC. I use v4.1 kernel released by fsl-community
with patches from VIA Tech. Inc. for VAB820, and my backport patches for
relevant features for this patchset. I use Ubuntu 17.04 from
ports.ubuntu.com as user land for armhf architecture.
$ aplay -v -M -D hw:imx6vab820sgtl5,0 /dev/urandom -f S16_LE -r 48000 --period-size=128 --buffer-size=256
Playing raw data '/dev/urandom' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono
Hardware PCM card 0 'imx6-vab820-sgtl5000' device 0 subdevice 0
Its setup is:
stream : PLAYBACK
access : MMAP_INTERLEAVED
format : S16_LE
subformat : STD
channels : 1
rate : 48000
exact rate : 48000 (48000/1)
msbits : 16
buffer_size : 256
period_size : 128
period_time : 2666
tstamp_mode : NONE
tstamp_type : MONOTONIC
period_step : 1
avail_min : 128
period_event : 0
start_threshold : 256
stop_threshold : 256
silence_threshold: 0
silence_size : 0
boundary : 1073741824
appl_ptr : 0
hw_ptr : 0
mmap_area[0] = 0x76f98000,0,16 (16)
$ trace-cmd record -e snd_pcm:hwptr -e snd_pcm:applptr
$ trace-cmd report
...
60.208495: applptr: pcmC0D0p/sub0: prev=1792, curr=1792, avail=0, period=128, buf=256
60.208633: applptr: pcmC0D0p/sub0: prev=1792, curr=1792, avail=0, period=128, buf=256
60.210022: hwptr: pcmC0D0p/sub0: IRQ: pos=128, old=1536, base=1536, period=128, buf=256
60.210202: applptr: pcmC0D0p/sub0: prev=1792, curr=1792, avail=128, period=128, buf=256
60.210344: hwptr: pcmC0D0p/sub0: POS: pos=128, old=1664, base=1536, period=128, buf=256
60.210348: applptr: pcmC0D0p/sub0: prev=1792, curr=1792, avail=128, period=128, buf=256
60.210486: applptr: pcmC0D0p/sub0: prev=1792, curr=1792, avail=128, period=128, buf=256
60.210626: applptr: pcmC0D0p/sub0: prev=1792, curr=1920, avail=0, period=128, buf=256
60.211002: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=0, period=128, buf=256
60.211142: hwptr: pcmC0D0p/sub0: POS: pos=128, old=1664, base=1536, period=128, buf=256
60.211146: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=0, period=128, buf=256
60.211287: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=0, period=128, buf=256
60.212690: hwptr: pcmC0D0p/sub0: IRQ: pos=0, old=1664, base=1536, period=128, buf=256
60.212866: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=128, period=128, buf=256
60.212999: hwptr: pcmC0D0p/sub0: POS: pos=0, old=1792, base=1792, period=128, buf=256
60.213003: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=128, period=128, buf=256
60.213135: applptr: pcmC0D0p/sub0: prev=1920, curr=1920, avail=128, period=128, buf=256
60.213276: applptr: pcmC0D0p/sub0: prev=1920, curr=2048, avail=0, period=128, buf=256
60.213654: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=0, period=128, buf=256
60.213796: hwptr: pcmC0D0p/sub0: POS: pos=0, old=1792, base=1792, period=128, buf=256
60.213800: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=0, period=128, buf=256
60.213937: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=0, period=128, buf=256
60.215356: hwptr: pcmC0D0p/sub0: IRQ: pos=128, old=1792, base=1792, period=128, buf=256
60.215542: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=128, period=128, buf=256
60.215679: hwptr: pcmC0D0p/sub0: POS: pos=128, old=1920, base=1792, period=128, buf=256
60.215683: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=128, period=128, buf=256
60.215813: applptr: pcmC0D0p/sub0: prev=2048, curr=2048, avail=128, period=128, buf=256
60.215947: applptr: pcmC0D0p/sub0: prev=2048, curr=2176, avail=0, period=128, buf=256
...
We can surely see 'applptr' event is probed even if the application run
for mmap programming scenario ('-M' option and 'hw' plugin). Below is a
result of strace:
02:44:15.886382 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.887203 poll([{fd=4, events=POLLOUT|POLLERR|POLLNVAL}], 1, -1) = 1 ([{fd=4, revents=POLLOUT}])
02:44:15.887471 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.887637 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.887805 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.887969 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.888132 read(3, "..."..., 256) = 256
02:44:15.889040 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.889221 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.889431 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.889606 poll([{fd=4, events=POLLOUT|POLLERR|POLLNVAL}], 1, -1) = 1 ([{fd=4, revents=POLLOUT}])
02:44:15.889833 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.889998 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.890164 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.891048 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.891228 read(3, "..."..., 256) = 256
02:44:15.891497 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.891661 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.891829 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
02:44:15.891991 poll([{fd=4, events=POLLOUT|POLLERR|POLLNVAL}], 1, -1) = 1 ([{fd=4, revents=POLLOUT}])
02:44:15.893007 ioctl(4, SNDRV_PCM_IOCTL_SYNC_PTR, 0x56a32b30) = 0
We can see 7 calls of ioctl(2) with SNDRV_PCM_IOCTL_SYNC_PTR per loop with
call of poll(2). 128 PCM frames are transferred per loop of one poll(2),
because the PCM substream is configured with S16_LE format and 1 channel
(2 byte * 1 * 128 = 256 bytes). This equals to the size of period of PCM
buffer. Comparing to the probed data, one of the 7 calls of ioctl(2) is
actually used to commit the number of copied PCM frames to kernel space.
The other calls are just used to check runtime status of PCM substream;
e.g. XRUN.
The tracepoint event is useful to investigate this case. I note that below
modules are related to the above sample.
* snd-soc-dummy.ko
* snd-soc-imx-sgtl5000.ko
* snd-soc-fsl-ssi.ko
* snd-soc-imx-pcm-dma.ko
* snd-soc-sgtl5000.ko
My additional note is lock acquisition. The event is probed under acquiring
PCM stream lock. This means that calculation in the event is free from
any hardware events.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In a series of recent work, ALSA PCM core got some arrangements to handle
application-side position on PCM buffer. However, relevant codes still
disperse to two translation units
This commit unifies these codes into a helper function.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Many drivers bind the sequencer stuff in off-load by another driver
module, so that it's loaded only on demand. In the current code, this
mechanism doesn't work when the driver is built-in while the sequencer
is module. We check with IS_REACHABLE() and enable only when the
sequencer is in the same level of build.
However, this is basically a overshoot. The binder code
(snd-seq-device) is an individual module from the sequencer core
(snd-seq), and we just have to make the former a built-in while
keeping the latter a module for allowing the scenario like the above.
This patch achieves that by rewriting Kconfig slightly. Now, a driver
that provides the manual sequencer device binding should select
CONFIG_SND_SEQ_DEVICE in a way as
select SND_SEQ_DEVICE if SND_SEQUENCER != n
Note that the "!=n" is needed here to avoid the influence of the
sequencer core is module while the driver is built-in.
Also, since rawmidi.o may be linked with snd_seq_device.o when
built-in, we have to shuffle the code to make the linker happy.
(the kernel linker isn't smart enough yet to handle such a case.)
That is, snd_seq_device.c is moved to sound/core from sound/core/seq,
as well as Makefile.
Last but not least, the patch replaces the code using IS_REACHABLE()
with IS_ENABLED(), since now the condition meets always when enabled.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The ppmu_events array is accessed only in this compilation unit so it
can be made static.
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Switch to use managed variant of acpi_dev_add_driver_gpios() to simplify
error path and fix potentially wrong assingment if ->probe() fails.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
The implementation details for SCPI seems to suggest that the sensor
readings must be reported by SCP using a well defined scale
(millidegree Celsius for temperature, millivolts for voltage,
milliamperes for current, microwatts for power and microjoules for
energy).
This is also important for the interaction with other subsystems: for
example both the thermal sub-system and the hwmon sysfs interface expect
the temperature expressed in millidegree Celsius.
Unfortunately since this behaviour is dependent on the firmware
implementation there are cases where the sensor readings are reported
using a different scale. For example in the Amlogic SoCs the
temperature is reported in degree and not millidegree Celsius.
To take into account this discrepancy and fixup the values reported by
SCP a new compatible 'amlogic,meson-gxbb-scpi-sensors' is introduced and
used in this patch by the scpi-hwmon driver to convert the sensor
readings to the expected scale.
Signed-off-by: Carlo Caione <carlo@endlessm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
The aspeed-pwm-tacho controller supports measuring the fan tach by using
leading, falling, or both edges. This change allows the driver to
support either of the three configurations and will appropriately modify
the returned tach data.
If the controller is measuring with both edges it can return a value more
quickly to the requestor. This version of the driver should still take ~1s
to return with an RPM value per fan, however, it can be tuned faster with
double edge counting enabled than without.
I tested this and found the number returned matched what I expected.
Signed-off-by: Patrick Venture <venture@google.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
OCC provides historical minimum and maximum value for the sensor
readings. This patch exports them as highest and lowest attributes
for the inband sensors copied by OCC to main memory.
Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
include/linux/i2c is not for client devices. Move the header file to a
more appropriate location.
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
When enabled temperature smoothing allows ramping the fan speed over a
configurable period of time instead of jumping to the new speed
instantaneously.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Systems using 4-wire fans usually require high frequency (22.5kHz)
output on the pwm. Add 22500 as a valid option in the pwmfreq_table. In
high frequency mode the low-order bit are ignored so they can safely be
set to 0.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
By default adt7475 will stop the fans (pwm duty cycle 0%) when the
temperature drops past Tmin - hysteresis. Some systems want to keep the
fans moving even when the temperature drops so add new sysfs attributes
that configure the enhanced acoustics min 1-3 which allows the fans to
run at the minimum configure pwm duty cycle.
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
The adt7475 has had find_nearest() since it's creation in 2009. Since
then find_closest() has been introduced and several drivers have been
updated to use it. Update the adt7475 to use find_closest() and remove
the now unused find_nearest().
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
IR35221 is a Digital DC-DC Multiphase Converter
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
[groeck: Preserve alphabetic order in Kconfig;
add missing break statements (from Dan Carpenter);
add missing error checks]
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Recent chips support multiple pins for fan speed inputs and fan control
outputs. Examine all of them to determine supported fan controls.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Instead of checking if a temperature source has a label, use a bit mask
to determine if a temperature source is valid for a given chip.
This simplifies the code and, if necessary, lets us support chips with
unknown or incomplete labels.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Pull key subsystem fixes from James Morris:
"Here are a bunch of fixes for Linux keyrings, including:
- Fix up the refcount handling now that key structs use the
refcount_t type and the refcount_t ops don't allow a 0->1
transition.
- Fix a potential NULL deref after error in x509_cert_parse().
- Don't put data for the crypto algorithms to use on the stack.
- Fix the handling of a null payload being passed to add_key().
- Fix incorrect cleanup an uninitialised key_preparsed_payload in
key_update().
- Explicit sanitisation of potentially secure data before freeing.
- Fixes for the Diffie-Helman code"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security: (23 commits)
KEYS: fix refcount_inc() on zero
KEYS: Convert KEYCTL_DH_COMPUTE to use the crypto KPP API
crypto : asymmetric_keys : verify_pefile:zero memory content before freeing
KEYS: DH: add __user annotations to keyctl_kdf_params
KEYS: DH: ensure the KDF counter is properly aligned
KEYS: DH: don't feed uninitialized "otherinfo" into KDF
KEYS: DH: forbid using digest_null as the KDF hash
KEYS: sanitize key structs before freeing
KEYS: trusted: sanitize all key material
KEYS: encrypted: sanitize all key material
KEYS: user_defined: sanitize key payloads
KEYS: sanitize add_key() and keyctl() key payloads
KEYS: fix freeing uninitialized memory in key_update()
KEYS: fix dereferencing NULL payload with nonzero length
KEYS: encrypted: use constant-time HMAC comparison
KEYS: encrypted: fix race causing incorrect HMAC calculations
KEYS: encrypted: fix buffer overread in valid_master_desc()
KEYS: encrypted: avoid encrypting/decrypting stack buffers
KEYS: put keyring if install_session_keyring_to_cred() fails
KEYS: Delete an error message for a failed memory allocation in get_derived_key()
...
Commit abb2ea7dfd ("compiler, clang: suppress warning for unused
static inline functions") just caused more warnings due to re-defining
the 'inline' macro.
So undef it before re-defining it, and also add the 'notrace' attribute
like the gcc version that this is overriding does.
Maybe this makes clang happier.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make use of recently implemented stack_depth tracking for arm64 JIT,
so that stack usage can be reduced heavily for programs not using
tail calls at least.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes two issues:
1) When forwarding on *,G mroutes that are in a vrf, the
kernel was dropping information about the actual incoming
interface when calling ip_mr_forward from ip_mr_input.
This caused ip_mr_forward to send the multicast packet
back out the incoming interface. Fix this by
modifying ip_mr_forward to be handed the correctly
resolved dev.
2) When a unresolved cache entry is created we store
the incoming skb on the unresolved cache entry and
upon mroute resolution from the user space daemon,
we attempt to forward the packet. Again we were
not resolving to the correct incoming device for
a vrf scenario, before calling ip_mr_forward.
Fix this by resolving to the correct interface
and calling ip_mr_forward with the result.
Fixes: e58e415968 ("net: Enable support for VRF with ipv4 multicast")
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Acked-by: David Ahern <dsahern@gmail.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Reviewed-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
mlx5-updates-2017-06-11
This series provides updates to mlx5 header rewrite feature, from Or Gerlitz.
and three more small updates From maor and eran.
-------
Or says:
Packets belonging to flows which are different by matching may still need
to go through the same header re-writes (e.g set the current routing hop
MACs and issue TTL decrement). To minimize the number of modify header
IDs, we add a cache for header re-write IDs which is keyed by the binary
chain of modify header actions.
The caching is supported for both eswitch and NIC use-cases, where the
actual conversion of the code to use caching comes in separate patches,
one per use-case.
Using a per field mask field, the TC pedit action supports modifying
partial fields. The last patch enables offloading that.
-------
From Maor, update flow table commands layout to the latest HW spec.
From Eran, ethtool connector type reporting updates.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
Mellanox mlx5 fixes 2017-06-11
This series contains some fixes for the mlx5 core and netdev driver.
Please pull and let me know if there's any problem.
For -stable:
("net/mlx5e: Added BW check for DIM decision mechanism") kernels >= 4.9
("net/mlx5e: Fix wrong indications in DIM due to counter wraparound") kernels >= 4.9
("net/mlx5: Remove several module events out of ethtool stats") kernels >= 4.10
("net/mlx5: Enable 4K UAR only when page size is bigger than 4K") kernels >= 4.11
*all patches apply with no issue on their -stable.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Netanel Belgazal says:
====================
Bugs fixes in ena ethernet driver
This patchset contains fixes for the bugs that were discovered so far.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
check_for_missing_tx_completions() is called from a timer
task and looking for lost tx packets.
The old implementation accumulate all the lost tx packets
and did not check if those packets were retrieved on a later stage.
This cause to a situation where the driver reset
the device for no reason.
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the rare case where the device runs out of free rx buffer
descriptors (in case of pressure on kernel memory),
and the napi handler continuously fail to refill new Rx descriptors
until device rx queue totally runs out of all free rx buffers
to post incoming packet, leading to a deadlock:
* The device won't send interrupts since all the new
Rx packets will be dropped.
* The napi handler won't try to allocate new Rx descriptors
since allocation is part of NAPI that's not being invoked any more
The fix involves detecting this scenario and rescheduling NAPI
(to refill buffers) by the keepalive/watchdog task.
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch also change the mapping functions to devm_ functions
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bug:
"Completion context is occupied" error printout will be noticed in
dmesg.
This error will cause the admin command to fail, which will lead to
an ena_probe() failure or a watchdog reset (depends on which admin
command failed).
Root cause:
__ena_com_submit_admin_cmd() is the function that submits new entries to
the admin queue.
The function have a check that makes sure the queue is not full and the
function does not override any outstanding command.
It uses head and tail indexes for this check.
The head is increased by ena_com_handle_admin_completion() which runs
from interrupt context, and the tail index is increased by the submit
function (the function is running under ->q_lock, so there is no risk
of multithread increment).
Each command is associated with a completion context. This context
allocated before call to __ena_com_submit_admin_cmd() and freed by
ena_com_wait_and_process_admin_cq_interrupts(), right after the command
was completed.
This can lead to a state where the head was increased, the check passed,
but the completion context is still in use.
Solution:
Use the atomic variable ->outstanding_cmds instead of using the head and
the tail indexes.
This variable is safe for use since it is bumped in get_comp_ctx() in
__ena_com_submit_admin_cmd() and is freed by comp_ctxt_release()
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixing a bug that the driver does not unmask the IO interrupts
in ndo_open():
occasionally, the MSI-X interrupt (for one or more IO queues)
can be masked when ndo_close() was called.
If that is followed by ndo open(),
then the MSI-X will be still masked so no interrupt
will be received by the driver.
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current flow to detect admin completion is:
while (command_not_completed) {
if (timeout)
error
check_for_completion()
sleep()
}
So in case the sleep took more than the timeout
(in case the thread/workqueue was not scheduled due to higher priority
task or prolonged VMexit), the driver can detect a stall even if
the completion is present.
The fix changes the order of this function to first check for
completion and only after that check if the timeout expired.
Fixes: 1738cd3ed3 ("Add a driver for Amazon Elastic Network Adapters (ENA)")
Signed-off-by: Netanel Belgazal <netanel@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull randomness fixes from Ted Ts'o:
"Improve performance by using a lockless update mechanism suggested by
Linus, and make sure we refresh per-CPU entropy returned get_random_*
as soon as the CRNG is initialized"
* tag 'random_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/random:
random: invalidate batched entropy after crng init
random: use lockless method of accessing and updating f->reg_idx
Pull ext4 fixes from Ted Ts'o:
"Fix various bug fixes in ext4 caused by races and memory allocation
failures"
* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
ext4: fix fdatasync(2) after extent manipulation operations
ext4: fix data corruption for mmap writes
ext4: fix data corruption with EXT4_GET_BLOCKS_ZERO
ext4: fix quota charging for shared xattr blocks
ext4: remove redundant check for encrypted file on dio write path
ext4: remove unused d_name argument from ext4_search_dir() et al.
ext4: fix off-by-one error when writing back pages before dio read
ext4: fix off-by-one on max nr_pages in ext4_find_unwritten_pgoff()
ext4: keep existing extra fields when inode expands
ext4: handle the rest of ext4_mb_load_buddy() ENOMEM errors
ext4: fix off-by-in in loop termination in ext4_find_unwritten_pgoff()
ext4: fix SEEK_HOLE
jbd2: preserve original nofs flag during journal restart
ext4: clear lockdep subtype for quota files on quota off
