| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When we find a timewait connection in __inet_hash_connect() and reuse
it for a new connection request, we have a race window, releasing bind
list lock and reacquiring it in __inet_twsk_kill() to remove timewait
socket from list.
Another thread might find the timewait socket we already chose, leading to
list corruption and crashes.
Fix is to remove timewait socket from bind list before releasing the bind lock.
Note: This problem happens if sysctl_tcp_tw_reuse is set.
Reported-by: kapil dakhane <kdakhane@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
First patch changes __inet_hash_nolisten() and __inet6_hash()
to get a timewait parameter to be able to unhash it from ehash
at same time the new socket is inserted in hash.
This makes sure timewait socket wont be found by a concurrent
writer in __inet_check_established()
Reported-by: kapil dakhane <kdakhane@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
| |
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Here is an updated version, because ixgbe_get_ethtool_stats()
needs to call dev_get_stats() or "ethtool -S" wont give
correct tx_bytes/tx_packets values.
Several cpus can update netdev->stats.tx_bytes & netdev->stats.tx_packets
in parallel. In this case, TX stats are under estimated and false sharing
takes place.
After a pktgen session sending exactly 200000000 packets :
# ifconfig fiber0 | grep TX
TX packets:198501982 errors:0 dropped:0 overruns:0 carrier:0
Multi queue devices should instead use txq->tx_bytes & txq->tx_packets
in their xmit() method (appropriate txq lock already held by caller, no
cache line miss), or use appropriate locking.
After patch, same pktgen session gives :
# ifconfig fiber0 | grep TX
TX packets:200000000 errors:0 dropped:0 overruns:0 carrier:0
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
A workaround added for all ESB2 devices (adds a delay for all MDIC accesses
which resolves an issue with the MDIC ready bit being set prematurely) is
applicable only to devices in which the MAC-PHY interconnect is not
operating in a certain mode with in-band MDIO. Check the control register
for the operating mode and enable the workaround accordingly.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The GG82563_REG() macro should not be used to determine the offset provided
to the e1000e_[read|write]_kmrn_reg() functions since the first argument to
the macro is already implied and gets masked off anyway in the functions.
The resultant register reads/writes with this patch are functionally the
same as before.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
| |
Bit 7 in the CTRL_REG register is actually the Software Definable Pin 3,
not the Software Definable Pin 7.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |
|
|
|
|
|
|
|
| |
GCC even warns about it, as reported by Andrew Morton:
net/ipv4/tcp.c: In function 'do_tcp_getsockopt':
net/ipv4/tcp.c:2544: warning: comparison is always false due to limited range of data type
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| |\
| |
| |
| | |
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next-2.6
|
| | |
| |
| |
| |
| |
| |
| | |
Adds support for the WiFi activity LED on the Dell Vostro A860 laptop.
Signed-off-by: Shahar Or <shahar@shahar-or.co.il>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| | |
A KERN_DEBUG didn't get removed when transitioning from printk to
pr_debug
Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There are different bits used to convey the setting of the rfkill
switch to the driver. The current driver only supports one of these
possibilities. These changes were derived from the latest version
of the vendor driver.
This patch fixes the regression noted in kernel Bugzilla #14743.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Reported-and-tested-by: Antti Kaijanmäki <antti@kaijanmaki.net>
Tested-by: Hin-Tak Leung <hintak.leung@gmail.com>
Cc: Stable <stable@kernel.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch fixes a bug in ath9k's tx status check, which
caused mac80211 to consider regularly transmitted unicast frames
as un-acked.
When checking the ts_status field for errors, it needs to be masked
with ATH9K_TXERR_FILT, because this field also contains other fields
like ATH9K_TX_ACKED.
Without this patch, AP mode is pretty much unusable, as hostapd
checks the ACK status for the frames that it injects.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
On a 32-bit machine, BIT() macro does not give the required
bit value if the bit is mroe than 31. In ieee802_11_parse_elems_crc(),
BIT() is suppossed to get the bit value more than 31 (42 (id of ERP_INFO_IE),
37 (CHANNEL_SWITCH_IE), (42), 32 (POWER_CONSTRAINT_IE), 45 (HT_CAP_IE),
61 (HT_INFO_IE)). As we do not get the required bit value for the above
IEs, crc over these IEs are never calculated, so any dynamic change in these
IEs after the association is not really handled on 32-bit platforms.
This patch fixes this issue.
Cc: stable@kernel.org
Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
net/rfkill/core.c: In function 'rfkill_type_show':
net/rfkill/core.c:610: warning: control may reach end of non-void function 'rfkill_get_type_str' being inlined
A gcc bug, but simple enough to squish.
Cc: John W. Linville <linville@tuxdriver.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
The FINALIZE_JOIN firmware command only looks at the first couple of
fields in the beacon, and therefore it's not necessary to complain if
the beacon is longer than 128 bytes.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When disassociating, mac80211 zeroes vif->bss_info.bssid before
calling our ->bss_info_changed(), but we need the BSSID to remove the
hardware station database entry for our AP, so we can't clear our
local copy of the BSSID until after we've done that.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| | |
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| | |
Don't forget to call pci_disable_device() if pci_request_regions()
fails during probe.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The time between loading the helper image and starting to upload the
main firmware image should be at least 5 ms or so. We were doing an
msleep(1) before, and 1 ms appears to not be enough in almost all
cases, but building with HZ=100 has always masked this so far. Bumping
the msleep argument to 5 fixes firmware loading e.g. when HZ=1000.
Some firmware images need more than 200ms to initialize. Bump the
ready code timeout to 500ms to accommodate for this.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Before issuing any firmware commands, we wait for the transmit rings
to drain, to prevent control versus data path synchronization issues.
In some cases, this can end up taking longer than the current hardcoded
limit of 5 seconds, for example if the transmit rings are filled with
packets for a host that has dropped off the air and we end up
retransmitting every pending packet at the lowest rate a couple of
times.
This patch changes mwl8k_tx_wait_empty() to only bail out on timeout
expiry if there was no change in the number of packets pending in the
transmit rings during the waiting period. If at least one transmit
ring entry was reclaimed while we were waiting, we are apparently still
making progress, and we'll allow waiting for another timeout period.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
Some firmware commands can under some circumstances take more than 2
seconds to complete. This patch bumps the timeout up to 10 seconds,
and prints a message whenever a command takes more than 2 seconds.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
On 8366, bit 6 in the rx descriptor rate field indicates whether the
packet was received on a 20MHz or 40MHz channel, and is not part of
the MCS index. Handle this properly, which then prevents hitting the
WARN_ON and being dropped in ieee80211_rx().
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When inserting a DMA header into a packet for transmission,
mwl8k_add_dma_header() would blindly zero the addr4 field, which
is not a good idea if the packet being transmitted is actually a
4-address packet.
Also, if the transmitted packet was a 4-address with QoS packet,
the memmove() to do the needed header reshuffling would inadvertently
overwrite the first two bytes of the packet payload with the QoS field.
This fixes both of these issues.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Packets exchanged between the mwl8k driver and the firmware always
have a 4-address header without QoS field. For QoS packets, the QoS
field is passed to/from the firmware via the tx/rx descriptors.
We were handling this correctly on transmit, but not on receive -- if
a QoS packet was received, we would leave garbage in the QoS field in
the packet passed up to the stack, which is Bad(tm).
Also, if the packet received on the air was a 4-address without QoS
packet, we would forget to skb_pull the 2-byte DMA length prefix off.
This patch adds an argument to the ->rxd_process() receive descriptor
operation to retrieve the QoS field from the receive descriptor, and
extends mwl8k_remove_dma_header() to insert this field back into the
packet if the packet received is a QoS packet. It also fixes
mwl8k_remove_dma_header() to strip off the length prefix in all cases.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There exist 12 802.11b/g rates, but mwl8k supports two additional
(non-standard) rates, and includes those rates in rate bitmasks and
in its internal rate table that hardware rate indices index.
Commit "mwl8k: report rate and other information for received frames"
added one of the nonstandard rates to the mwl8k_rates table to make
the OFDM rates in the table line up with the rate indices that are
reported in the receive descriptor (so that we can just simply copy
the receive descriptor rate index into ieee80211_rx_status::rate_idx)
and bumped MWL8K_IEEE_LEGACY_DATA_RATES from 12 to 13, but this
screwed up the UPDATE_STADB command struct layout, as it also uses
that define, for its legacy_rates array.
To avoid having to convert rate indices and legacy rate bitmaps (e.g.
ieee80211_bss_conf::basic_rates) between the 12-rate mac80211 format
and the 14-rate mwl8k format, we'll report all 14 rates in our wiphy's
band, but filter out the nonstandard ones e.g. in the case of the
UPDATE_STADB command which only accepts 12 rates.
In the commands that accept 14 rates (SET_AID, SET_RATE), replace the
use of the MWL8K_RATE_INDEX_MAX_ARRAY define in the command struct by
the constant 14, to make it clearer that these commands accept 14 rates.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The MCS bitmaps in the SET_RATE command structure were of the wrong
size, due to use of the wrong define for the array length. Just
hardcode the lengths as 16, and do the same for the MCS bitmaps in
other command structures.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Not only ps_sdata but also IEEE80211_CONF_PS is to be considered
before restoring PS in scan_ps_disable(). For instance, when ps_sdata
is set but CONF_PS is not set just because the dynamic timer is still
running, a sw scan leads to setting of CONF_PS in scan_ps_disable
instead of restarting the dynamic PS timer.
Also for the above case, a null data frame is to be sent after
returning to operating channel which was not happening with the
current implementation. This patch fixes this too.
Signed-off-by: Vivek Natarajan <vnatarajan@atheros.com>
Reviewed-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
hwsim testing has revealed that when the MLME
recalculates the idle state of the device, it
sometimes does so before sending the final
deauthentication or disassociation frame. This
patch changes the place where the idle state
is recalculated, but of course driver transmit
is typically asynchronous while configuration
is expected to be synchronous, so it doesn't
fix all possible cases yet.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| | |
The bssid can be zero when null data template is set in wl1251_op_config().
It's enough, and especially safe, to set it once after association.
Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
bssid needs to be copied first in wl1251_op_bss_info_changed(), otherwise
templates will have incorrect bssid and power save will not work
correctly.
Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There was a warning from wl1251_op_bss_info_changed():
wl1251: WARNING Set ctsprotect failed 0
It was printed always, it's completely false and can be removed.
Signed-off-by: Kalle Valo <kalle.valo@nokia.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
CC [M] drivers/net/wireless/iwlwifi/iwl-tx.o
drivers/net/wireless/iwlwifi/iwl-tx.c: In function ‘iwl_tx_agg_stop’:
drivers/net/wireless/iwlwifi/iwl-tx.c:1356: warning: passing argument 1 of ‘ieee80211_stop_tx_ba_cb_irqsafe’ from incompatible pointer type
include/net/mac80211.h:2128: note: expected ‘struct ieee80211_vif *’ but argument is of type ‘struct ieee80211_hw *’
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
I messed up the merge in d7fc02c7bae7b1cf69269992cf880a43a350cdaa, where
the conflict in question wasn't just about CTL_UNNUMBERED being removed,
but the 'strategy' field is too (sysctl handling is now done through the
/proc interface, with no duplicate protocols for reading the data).
Reported-by: Larry Finger <Larry.Finger@lwfinger.net>
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
| |\ \
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* 'for-2.6.33' of git://git.kernel.dk/linux-2.6-block: (113 commits)
cfq-iosched: Do not access cfqq after freeing it
block: include linux/err.h to use ERR_PTR
cfq-iosched: use call_rcu() instead of doing grace period stall on queue exit
blkio: Allow CFQ group IO scheduling even when CFQ is a module
blkio: Implement dynamic io controlling policy registration
blkio: Export some symbols from blkio as its user CFQ can be a module
block: Fix io_context leak after failure of clone with CLONE_IO
block: Fix io_context leak after clone with CLONE_IO
cfq-iosched: make nonrot check logic consistent
io controller: quick fix for blk-cgroup and modular CFQ
cfq-iosched: move IO controller declerations to a header file
cfq-iosched: fix compile problem with !CONFIG_CGROUP
blkio: Documentation
blkio: Wait on sync-noidle queue even if rq_noidle = 1
blkio: Implement group_isolation tunable
blkio: Determine async workload length based on total number of queues
blkio: Wait for cfq queue to get backlogged if group is empty
blkio: Propagate cgroup weight updation to cfq groups
blkio: Drop the reference to queue once the task changes cgroup
blkio: Provide some isolation between groups
...
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fix a crash during boot reported by Jeff Moyer. Fix the issue of accessing
cfqq after freeing it.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@carl.(none)>
|
| | | |
| | |
| | |
| | |
| | | |
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
After the merge of the IO controller patches, booting on my megaraid
box ran much slower. Vivek Goyal traced it down to megaraid discovery
creating tons of devices, each suffering a grace period when they later
kill that queue (if no device is found).
So lets use call_rcu() to batch these deferred frees, instead of taking
the grace period hit for each one.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
o Now issues of blkio controller and CFQ in module mode should be fixed.
Enable the cfq group scheduling support in module mode.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
o One of the goals of block IO controller is that it should be able to
support mulitple io control policies, some of which be operational at
higher level in storage hierarchy.
o To begin with, we had one io controlling policy implemented by CFQ, and
I hard coded the CFQ functions called by blkio. This created issues when
CFQ is compiled as module.
o This patch implements a basic dynamic io controlling policy registration
functionality in blkio. This is similar to elevator functionality where
ioschedulers register the functions dynamically.
o Now in future, when more IO controlling policies are implemented, these
can dynakically register with block IO controller.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
o blkio controller is inside the kernel and cfq makes use of interfaces
exported by blkio. CFQ can be a module too, hence export symbols used
by CFQ.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
With CLONE_IO, parent's io_context->nr_tasks is incremented, but never
decremented whenever copy_process() fails afterwards, which prevents
exit_io_context() from calling IO schedulers exit functions.
Give a task_struct to exit_io_context(), and call exit_io_context() instead of
put_io_context() in copy_process() cleanup path.
Signed-off-by: Louis Rilling <louis.rilling@kerlabs.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
With CLONE_IO, copy_io() increments both ioc->refcount and ioc->nr_tasks.
However exit_io_context() only decrements ioc->refcount if ioc->nr_tasks
reaches 0.
Always call put_io_context() in exit_io_context().
Signed-off-by: Louis Rilling <louis.rilling@kerlabs.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
cfq_arm_slice_timer() has logic to disable idle window for SSD device. The same
thing should be done at cfq_select_queue() too, otherwise we will still see
idle window. This makes the nonrot check logic consistent in cfq.
Tests in a intel SSD with low_latency knob close, below patch can triple disk
thoughput for muti-thread sequential read.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | | |
It's currently not an allowed configuration, so express that in Kconfig.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
They should not be declared inside some other file that's not related
to CFQ.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | | |
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | | |
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
o rq_noidle() is supposed to tell cfq that do not expect a request after this
one, hence don't idle. But this does not seem to work very well. For example
for direct random readers, rq_noidle = 1 but there is next request coming
after this. Not idling, leads to a group not getting its share even if
group_isolation=1.
o The right solution for this issue is to scan the higher layers and set
right flag (WRITE_SYNC or WRITE_ODIRECT). For the time being, this single
line fix helps. This should not have any significant impact when we are
not using cgroups. I will later figure out IO paths in higher layer and
fix it.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
o If a group is running only a random reader, then it will not have enough
traffic to keep disk busy and we will reduce overall throughput. This
should result in better latencies for random reader though. If we don't
idle on random reader service tree, then this random reader will experience
large latencies if there are other groups present in system with sequential
readers running in these.
o One solution suggested by corrado is that by default keep the random readers
or sync-noidle workload in root group so that during one dispatch round
we idle only once on sync-noidle tree. This means that all the sync-idle
workload queues will be in their respective group and we will see service
differentiation in those but not on sync-noidle workload.
o Provide a tunable group_isolation. If set, this will make sure that even
sync-noidle queues go in their respective group and we wait on these. This
provides stronger isolation between groups but at the expense of throughput
if group does not have enough traffic to keep the disk busy.
o By default group_isolation = 0
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
