| Age | Commit message (Collapse) | Author |
|---|
|
Dedicated small file workloads have been seeing significant free
space fragmentation causing premature inode allocation failure
when large inode sizes are in use. A particular test case showed
that a workload that runs to a real ENOSPC on 256 byte inodes would
fail inode allocation with ENOSPC about about 80% full with 512 byte
inodes, and at about 50% full with 1024 byte inodes.
The same workload, when run with -o allocsize=4096 on 1024 byte
inodes would run to being 100% full before giving ENOSPC. That is,
no freespace fragmentation at all.
The issue was caused by the specific IO pattern the application had
- the framework it was using did not support direct IO, and so it
was emulating it by using fadvise(DONT_NEED). The result was that
the data was getting written back before the speculative prealloc
had been trimmed from memory by the close(), and so small single
block files were being allocated with 2 blocks, and then having one
truncated away. The result was lots of small 4k free space extents,
and hence each new 8k allocation would take another 8k from
contiguous free space and turn it into 4k of allocated space and 4k
of free space.
Hence inode allocation, which requires contiguous, aligned
allocation of 16k (256 byte inodes), 32k (512 byte inodes) or 64k
(1024 byte inodes) can fail to find sufficiently large freespace and
hence fail while there is still lots of free space available.
There's a simple fix for this, and one that has precendence in the
allocator code already - don't do speculative allocation unless the
size of the file is larger than a certain size. In this case, that
size is the minimum default preallocation size:
mp->m_writeio_blocks. And to keep with the concept of being nice to
people when the files are still relatively small, cap the prealloc
to mp->m_writeio_blocks until the file goes over a stripe unit is
size, at which point we'll fall back to the current behaviour based
on the last extent size.
This will effectively turn off speculative prealloc for very small
files, keep preallocation low for small files, and behave as it
currently does for any file larger than a stripe unit. This
completely avoids the freespace fragmentation problem this
particular IO pattern was causing.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
|
|
Similar to bulkstat inode chunk readahead, we need to plug directory
data buffer readahead during getdents to ensure that we can merge
adjacent readahead requests and sort out of order requests optimally
before they are dispatched. This improves the readahead efficiency
and reduces the IO load it generates as the IO patterns are
significantly better for both contiguous and fragmented directories.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
|
|
I was running some tests on bulkstat on CRC enabled filesystems when
I noticed that all the IO being issued was 8k in size, regardless of
the fact taht we are issuing sequential 8k buffers for inodes
clusters. The IO size should be 16k for 256 byte inodes, and 32k for
512 byte inodes, but this wasn't happening.
blktrace showed that there was an explict plug and unplug happening
around each readahead IO from _xfs_buf_ioapply, and the unplug was
causing the IO to be issued immediately. Hence no opportunity was
being given to the elevator to merge adjacent readahead requests and
dispatch them as a single IO.
Add plugging around the inode chunk readahead dispatch loop in
bulkstat to ensure that we don't unplug the queue between adjacent
inode buffer readahead IOs and so we get fewer, larger IO requests
hitting the storage subsystem for bulkstat.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
|
|
CC drivers/net/wireless/ath/ath10k/mac.o
drivers/net/wireless/ath/ath10k/mac.c: In function ‘chan_to_phymode’:
drivers/net/wireless/ath/ath10k/mac.c:229:3: warning: enumeration value ‘NL80211_CHAN_WIDTH_5’ not handled in switch [-Wswitch]
drivers/net/wireless/ath/ath10k/mac.c:229:3: warning: enumeration value ‘NL80211_CHAN_WIDTH_10’ not handled in switch [-Wswitch]
drivers/net/wireless/ath/ath10k/mac.c:247:3: warning: enumeration value ‘NL80211_CHAN_WIDTH_5’ not handled in switch [-Wswitch]
drivers/net/wireless/ath/ath10k/mac.c:247:3: warning: enumeration value ‘NL80211_CHAN_WIDTH_10’ not handled in switch [-Wswitch]
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
This is needed so the interface combination can still be
validated when CONFIG_MAC80211_MESH is not enabled.
Otherwise wiphy registration fails.
Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
In the code of the receive path some code was dealing with how
things were done in older kernels. Not really needed for an
upstream driver.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
In the receive path a spinlock is taken upon parsing the TLV signal
header. This moves to locking to the TLV handling functions where
it protects the data structures.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Trivial cleanup of debug messages.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Transmit packets needs to be tagged in order to receive a tx status
feedback from the firmware. Determine the tag in the netdev transmit
callback instead of determining the tag just before transfer to the
device. This reduces the number of exception flows and hence makes
the driver code simpler.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
DMA engine of some old SDIO host controllers require block size alignment for
data length of each scatterlist item. This patch introduces an intermediate
buffer list to support this kind of platform. It decreases the throughput
because of an extra memcpy in critical data path. So don't turn this on unless
it's necessary.
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Introduce a unified dongle backplane address preparation function
brcmf_sdio_addrprep to replace duplicate address prep code.
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Remove SDIO_REQ_ASYNC from brcmfmac since it is not being used.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Used NL80211_NUM_ACS to indicate the BCMC fifo used in the driver
which has the same value now, but it is a bad idea relying on that.
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
When getting a transmit packet from the networking layer simply
enqueue the packet unconditional and have it handled by the dequeue
worker. The transfer of the packet to the bus-specific driver part
is now done from one context.
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
This allows enabling support for extra hardware with just a module
param, without kernel/module recompilation.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Apparently HW doesn't require us to generate MMIC
for TKIP suite.
Each frame was 8 bytes longer than it should be
and some APs would drop frames that exceed 1520
bytes of 802.11 payload. This could be observed
during throughput tests or fragmented IP traffic.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Nonsense channel flags were being set.
Although it doesn't seem this was visible to the
user the patch makes sure that channel
availability won't be crippled in the future if
ath_common behaviour changes.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Irqs were not freed up correctly upon msi-x setup
failure.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
The code writes the default_power2 value into the TX field
of the RFCSR50 register, however the condition in the if
statement uses default_power1. Due to this, wrong TX power
value might be written into the register.
Use the correct value in the condition to fix the issue.
Compile tested only.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Cc: stable@vger.kernel.org # 3.10
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
The 3T/3R devices are using the tertiary PAs/LNAs
however those are never turned on. Fix the code to
turn on those on for such devices.
Also modify the code to use switch statements to
improve readability.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
The secondary PAs/LNAs are turned on only for 2T/2R
devices, however these are used for 3T/3R devices as
well. Always turn those on if the device uses more
than one tx/rx chains.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
Ralink 3T chipsets are using a different EEPROM
layout than the others. The EEPROM on these devices
contain more data than the others which does not fit
into 272 byte which the rt2800 driver actually uses.
The Ralink reference driver defines EEPROM_SIZE to
512/1024 bytes for PCI/USB devices respectively.
Increase the EEPROM_SIZE constant to 512 bytes, in
order to make room for EEPROM data of 3T devices.
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Acked-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
NCT6775 does not support alarms for fans 4 and 5. Drop the attributes.
cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
Driver displays wrong alarms for temperature attributes.
Turns out that temperature alarm bits are not fixed, but determined
by temperature source mapping. To fix the problem, walk through
the temperature sources to determine the correct alarm bit associated
with a given attribute.
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
GMT G762/763 fan speed PWM controller is connected directly to a fan
and performs closed-loop or open-loop control of the fan speed. Two
modes - PWM or DC - are supported by the chip. Introduced driver
provides various knobs to control the operations of the chip (via
sysfs interface). Specific characteristics of the system can be passed
either using board init code or via DT. Documentation for both the
driver and DT bindings are also provided.
Signed-off-by: Arnaud Ebalard <arno@natisbad.org>
Tested-by: Simon Guinot <simon.guinot@sequanux.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
Adding another way that is device tree to pass the shunt resistor
value to driver except for platform data.
Signed-off-by: Tang Yuantian <Yuantian.Tang@freescale.com>
[Guenter Roeck: Added missing of.h include]
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
Replace some written information with tables to improve readability
and to simplify adding newer devices in the future.
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
These changes add DS1731 chip support to the ds1621 driver,
Kconfig, and documentation.
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
This helps the kernel to find the right module once the device is
created.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
LM84 does not support minimum temperature registers.
Only create the respective sysfs attributes for other chips.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>
|
|
Due to a lack of device and vendor identification registers, the
Dallas/Maxim DS16xx devices cannot be uniquely detected, sometimes
resulting in false positives. Therefore, the detect function is
being removed in favor of explicit device instantiation.
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
Add definitions, information, and code for ds1631 chip support
to the ds1621 driver.
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
The ds1721 device can be configured for 9..12 bit resolutions;
add a sysfs attribute for userspace to configure this attribute.
The definition, description, details, and usage are shown in the
documentation and were crafted from an LM73 driver patch done by
Chris Verges & Guenter Roeck).
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
Update the ds1621 documentation, driver, and Kconfig with
ds1721 chip support.
Signed-off-by: Robert Coulson <rob.coulson@gmail.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
|
|
If a GFS2 file system is mounted with quotas and a file is grown
in such a way that its free blocks for the allocation are represented
in a secondary bitmap, GFS2 ran out of blocks in the transaction.
That resulted in "fatal: assertion "tr->tr_num_buf <= tr->tr_blocks".
This patch reserves extra blocks for the quota change so the
transaction has enough space.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
|
|
Add platform specific functionality for the DW SD/MMC driver for
SoCFPGA. Move SDMMC_CMD_USE_HOLD_REG to dw_mmc.h so other platforms
can use this define.
Signed-off-by: Dinh Nguyen <dinguyen@altera.com>
Reviewed-by: Pavel Machek <pavel@denx.de>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
Although the HC supports HS200 (eMMC) the caps2 are always zero; this
means there's no way to use the super speed mode (when init the card).
If the HC support SDR104, for SD3.0, so it also supports HS200 for eMMC
and this patch just sets the MMC_CAP2_HS200 in the host caps2 field.
Reported-by: Youssef Triki <youssef.triki@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Reviewed-by: Philip Rakity <prakity@nvidia.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
Commit bb691ae464b77d30e74c66480e98d74e88d6b194 breaks boot on OLPC
XO-4, it hangs somewhere inside sdhci_add_host.
When pm_runtime_set_autosuspend_delay() was being called, the device's
usage counter was 0, causing the PM layer to runtime-suspend the
device. We then went on to call sdhci_add_host() on a suspended
device, which hung.
Fix this by making the driver consistent with the omap_hsmmc driver,
both in terms of runtime PM initialization and error handling. Now
the device is not runtime-suspended until we exit the probe routine.
Signed-off-by: Daniel Drake <dsd@laptop.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
The DT-binding for MMC_CAP2_FULL_PWR_CYCLE, is used to indicate whether
it is possible to perform a full power cycle of the card.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
MMC_CAP2_FULL_PWR_CYCLE shall be set by host drivers which are able to
do a complete power cycle of the card. In the eMMC case that includes
both vcc and vccq.
This CAP is providing the protocol layer with important information,
needed to take optimized decisions during card initialization and in
the suspend/resume sequence.
MMC_CAP2_POWEROFF_NOTIFY is replaced by MMC_CAP2_FULL_PWR_CYCLE, since
it makes sense to use a wider scope for it.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
In suspend mode it is important to save power. If the host is able to
cut buth vcc and vccq, the MMC_CAP2_POWEROFF_NOTIFY shall be set. It
will mean the card will be completely powered down at suspend and the
power off notification cmd will be sent prior power down.
It seems common not being able to cut both vcc and vccq for a host. In
this situation we issue the sleep cmd in favor of the power off
notification cmd, to save more power.
While maintainng the above policy, we also want to make use of the
power off notification in the shutdown sequence, even in the case were
the host has not set MMC_CAP2_POWEROFF_NOTIFY, since we know vcc and
vccq will regardless be cut.
We accomplish this by always enabling the power off notification byte
in the EXT_CSD and issue the power off notification when either
MMC_CAP2_POWEROFF_NOTIFY is set or we are executing a shutdown.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
The shutdown sequence of an (e)MMC is very similar to a suspend. We
re-use the suspend function and tell it we are not in suspend context.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
Depending on the context of the operation while powering down the card,
either POWER_OFF_NOTIFY_SHORT or POWER_OFF_NOTIFY_LONG will be used. In
suspend context a short timeout is preferred while a long timeout would
be acceptable in a shutdown/hibernation context.
We add a new parameter to the mmc_suspend function so we can provide an
indication of what notification type to use.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
For the SD .shutdown callback we re-use the SD suspend function since
it performs the relevant actions.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
By adding an optional .shutdown callback to the bus_ops struct we
provide the possibility to let each bus type handle it's shutdown
requirements.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
Considering shutdown of the card, the responsibility to initate this
sequence shall be driven from the mmc_bus.
This patch enables the mmc_bus to handle this sequence properly. A new
.shutdown callback is added in the mmc_driver struct which is used to
shutdown the blk device.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
The host should be responsible to suspend|resume the host and not the
card. This patch changes this behaviour, by moving the responsiblity
to the mmc bus instead which already holds the card device.
The exported functions mmc_suspend|resume_host are now to be considered
as depcrecated. Once all host drivers moves away from using them, we
can remove them. As of now, a successful error code is always returned.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
|
By moving code from the mmc_suspend|resume_host down into each
.suspend|resume bus_ops callback, we get a more flexible solution.
Some nice side effects are that we get a better understanding of each
bus_ops suspend|resume sequence and the common code don't have to take
care of specific corner cases, especially for the SDIO case.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
|
