| Age | Commit message (Collapse) | Author |
|---|
|
Sergei Shtylyov pointed out that two instances of parenthesis are not
needed, so remove them.
Suggested-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
|
|
wq_select_unbound_cpu() is designed for unbound workqueues only, but
it's wrongly called when using a bound workqueue too.
Fixing this ensures work queued to a bound workqueue with
cpu=WORK_CPU_UNBOUND always runs on the local CPU.
Before, that would happen only if wq_unbound_cpumask happened to include
it (likely almost always the case), or was empty, or we got lucky with
forced round-robin placement. So restricting
/sys/devices/virtual/workqueue/cpumask to a small subset of a machine's
CPUs would cause some bound work items to run unexpectedly there.
Fixes: ef557180447f ("workqueue: schedule WORK_CPU_UNBOUND work on wq_unbound_cpumask CPUs")
Cc: stable@vger.kernel.org # v4.5+
Signed-off-by: Hillf Danton <hdanton@sina.com>
[dj: massage changelog]
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
drm-intel-fixes
gvt-fixes-2020-03-10
- Fix vgpu idr destroy causing timer destroy failure (Zhenyu)
- Fix VBT size (Tina)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
From: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200310080933.GE28483@zhen-hp.sh.intel.com
|
|
Store the IOMMU mapping created by the device core of each Exynos DRM
sub-device and restore it when the Exynos DRM driver is unbound. This
fixes IOMMU initialization failure for the second time when a deferred
probe is triggered from the bind() callback of master's compound DRM
driver. This also fixes the following issue found using kmemleak
detector:
unreferenced object 0xc2137640 (size 64):
comm "swapper/0", pid 1, jiffies 4294937900 (age 3127.400s)
hex dump (first 32 bytes):
50 a3 14 c2 80 a2 14 c2 01 00 00 00 20 00 00 00 P........... ...
00 10 00 00 00 80 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<3acd268d>] arch_setup_dma_ops+0x4c/0x104
[<9f7d2cce>] of_dma_configure+0x19c/0x3a4
[<ba07704b>] really_probe+0xb0/0x47c
[<4f510e4f>] driver_probe_device+0x78/0x1c4
[<7481a0cf>] device_driver_attach+0x58/0x60
[<0ff8f5c1>] __driver_attach+0xb8/0x158
[<86006144>] bus_for_each_dev+0x74/0xb4
[<10159dca>] bus_add_driver+0x1c0/0x200
[<8a265265>] driver_register+0x74/0x108
[<e0f3451a>] exynos_drm_init+0xb0/0x134
[<db3fc7ba>] do_one_initcall+0x90/0x458
[<6da35917>] kernel_init_freeable+0x188/0x200
[<db3f74d4>] kernel_init+0x8/0x110
[<1f3cddf9>] ret_from_fork+0x14/0x20
[<8cd12507>] 0x0
unreferenced object 0xc214a280 (size 128):
comm "swapper/0", pid 1, jiffies 4294937900 (age 3127.400s)
hex dump (first 32 bytes):
00 a0 ec ed 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<3acd268d>] arch_setup_dma_ops+0x4c/0x104
[<9f7d2cce>] of_dma_configure+0x19c/0x3a4
[<ba07704b>] really_probe+0xb0/0x47c
[<4f510e4f>] driver_probe_device+0x78/0x1c4
[<7481a0cf>] device_driver_attach+0x58/0x60
[<0ff8f5c1>] __driver_attach+0xb8/0x158
[<86006144>] bus_for_each_dev+0x74/0xb4
[<10159dca>] bus_add_driver+0x1c0/0x200
[<8a265265>] driver_register+0x74/0x108
[<e0f3451a>] exynos_drm_init+0xb0/0x134
[<db3fc7ba>] do_one_initcall+0x90/0x458
[<6da35917>] kernel_init_freeable+0x188/0x200
[<db3f74d4>] kernel_init+0x8/0x110
[<1f3cddf9>] ret_from_fork+0x14/0x20
[<8cd12507>] 0x0
unreferenced object 0xedeca000 (size 4096):
comm "swapper/0", pid 1, jiffies 4294937900 (age 3127.400s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<3acd268d>] arch_setup_dma_ops+0x4c/0x104
[<9f7d2cce>] of_dma_configure+0x19c/0x3a4
[<ba07704b>] really_probe+0xb0/0x47c
[<4f510e4f>] driver_probe_device+0x78/0x1c4
[<7481a0cf>] device_driver_attach+0x58/0x60
[<0ff8f5c1>] __driver_attach+0xb8/0x158
[<86006144>] bus_for_each_dev+0x74/0xb4
[<10159dca>] bus_add_driver+0x1c0/0x200
[<8a265265>] driver_register+0x74/0x108
[<e0f3451a>] exynos_drm_init+0xb0/0x134
[<db3fc7ba>] do_one_initcall+0x90/0x458
[<6da35917>] kernel_init_freeable+0x188/0x200
[<db3f74d4>] kernel_init+0x8/0x110
[<1f3cddf9>] ret_from_fork+0x14/0x20
[<8cd12507>] 0x0
unreferenced object 0xc214a300 (size 128):
comm "swapper/0", pid 1, jiffies 4294937900 (age 3127.400s)
hex dump (first 32 bytes):
00 a3 14 c2 00 a3 14 c2 00 40 18 c2 00 80 18 c2 .........@......
02 00 02 00 ad 4e ad de ff ff ff ff ff ff ff ff .....N..........
backtrace:
[<08cbd8bc>] iommu_domain_alloc+0x24/0x50
[<b835abee>] arm_iommu_create_mapping+0xe4/0x134
[<3acd268d>] arch_setup_dma_ops+0x4c/0x104
[<9f7d2cce>] of_dma_configure+0x19c/0x3a4
[<ba07704b>] really_probe+0xb0/0x47c
[<4f510e4f>] driver_probe_device+0x78/0x1c4
[<7481a0cf>] device_driver_attach+0x58/0x60
[<0ff8f5c1>] __driver_attach+0xb8/0x158
[<86006144>] bus_for_each_dev+0x74/0xb4
[<10159dca>] bus_add_driver+0x1c0/0x200
[<8a265265>] driver_register+0x74/0x108
[<e0f3451a>] exynos_drm_init+0xb0/0x134
[<db3fc7ba>] do_one_initcall+0x90/0x458
[<6da35917>] kernel_init_freeable+0x188/0x200
[<db3f74d4>] kernel_init+0x8/0x110
[<1f3cddf9>] ret_from_fork+0x14/0x20
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
|
|
Jose Abreu says:
====================
net: Add support for Synopsys DesignWare XPCS
This adds support for Synopsys DesignWare XPCS in net subsystem and
integrates it into stmmac.
At 1/8, we start by removing the limitation of stmmac selftests that needed
a PHY to pass all the tests.
Then at 2/8 we use some helpers in stmmac so that some code can be
simplified.
At 3/8, we fallback to dev_fwnode() so that PCI based setups wich may
not have CONFIG_OF can still use FW node.
At 4/8, we adapt stmmac to the new PHYLINK changes as suggested by Russell
King.
We proceed by doing changes in PHYLINK in order to support XPCS: At 5/8 we
add some missing speeds that USXGMII supports and at 6/8 we check if
Autoneg is supported after initial parameters are validated.
Support for XPCS is finally introduced at 7/8, along with the usage of it
in stmmac driver at 8/8.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Adds all the necessary logic so that stmmac can be used with Synopsys
DesignWare XPCS.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Synopsys DesignWare XPCS is an MMD that can manage link status,
auto-negotiation, link training, ...
In this commit we add basic support for XPCS using USXGMII interface and
Clause 73 Auto-negotiation.
This is highly tied with PHYLINK and can't be used without it. A given
ethernet driver can use the provided callbacks to add the support for
XPCS.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We may have cases where MAC or PCS do not support Autoneg. Check if it
is supported after validate callback is called.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
USXGMII also supports these missing backplane speeds.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Convert the stmmac ethernet driver to use the finalised link parameters
in mac_link_up() rather than the parameters in mac_config().
Suggested-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When CONFIG_OF is not enabled, of_fwnode_handle() will return NULL, even
though we can have a FW handle from a given device.
Fallback to dev_fwnode() helper if needed.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use the linkmode_and()/linkmode_andnot() helpers to simplify the code.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If a PHY is not attached, we can still run the tests with MAC loopback
mode. Return -EOPNOTSUPP error code in PHY loopback test so that global
status is not a failure.
Signed-off-by: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5-updates-2020-03-09
This series provides updates to mlx5 driver:
1) Use vport metadata matching only when mandatory
2) Introduce root flow table and ethtool steering for uplink representors
3) Expose port speed via FW when link modes are not available
3) Misc cleanups
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Huazhong Tan says:
====================
net: hns3: misc updates for -net-next
This series includes some misc updates for the HNS3 ethernet driver.
[patch 1] fixes some mixed type operation warning.
[patch 2] renames a macro to make it more readable.
[patch 3 & 4] removes some unnecessary code.
[patch 5] adds check before assert VF reset to prevent some unsuitable
error log.
[patch 6 - 9] some modifications related to printing.
Change log:
V1->V2: fixes a wrong print format in [patch 1] suggested by Jian Shen.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Since kernel already has logs after kzalloc fails,
it's unnecessary to print duplicate logs. So this
patch deletes these logs.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch modifies some printing relating to reset issue.
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds a local variable to save the command code in
some dump cases which need to modify the command code, then
the failing command code can be print out for debugging.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When received an unknown vector 0 interrupt, there is not a
helpful information for user to realize that now. So this patch
prints out the value of the status register for this case, and
uses dev_info() instead of dev_dbg() in hclge_check_event_cause().
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When setting VF's MAC from PF, if the VF driver not loaded, the
firmware will return error to PF.
So PF should check whether VF is alive before sending message to
VF when setting VF's MAC.
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In hclge_add_mc_addr_common() and hclge_rm_mc_addr_common(),
variable req had been set as "0" by memset(), so it's unnecessary
to set .entry_type field as "0" with hnae3_set_bit() again.
Signed-off-by: Guojia Liao <liaoguojia@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In hclge_handle_hw_ras_error(), it is unnecessary to check
HCLGE_STATE_RST_HANDLING flag, because the reset priority
has been ensured by the process itself.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The name of macro HCLGE_MAX_NCL_CONFIG_LENGTH is inaccurate,
this patch renames it to HCLGE_NCL_CONFIG_LENGTH_IN_EACH_CMD.
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch cleans up some incorrect type in assignment reported by sparse
and compiler.
The warning as below:
- warning : restricted __le16 degrades to integer
- warning : cast from restricted __le32
- warning : cast from restricted __be32
- warning : cast from restricted __be16
and "mixed operation".
Signed-off-by: Guojia Liao <liaoguojia@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Shannon Nelson says:
====================
ionic updates
This is a set of small updates for the Pensando ionic driver, some
from internal work, some a result of mailing list discussions.
v4 - don't register mgmt device netdev
v3 - changed __attribute__(packed)) to __packed
v2 - removed print from ionic_init_module()
====================
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use the default kernel version in ethtool drv_info output
and drop the module version.
Cc: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for the management port device id. This is to
capture the device and set it up for devlink use, but not set
it up for network operations. We still use a netdev object
in order to use the napi infrasucture for processing adminq
and notifyq messages, we just don't register the netdev.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Print the PCI bus lane information so people can keep an
eye out for poor configurations that might not support
the advertised speed.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We can disable rxhashing by setting rss_types to 0. The user
can toggle this with "ethtool -K <ethX> rxhash off|on",
which calls into the .ndo_set_features callback with the
NETIF_F_RXHASH feature bit set or cleared. This patch adds
a check for that bit and updates the FW if necessary.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Remove the unused flags field and and fix the bitflag names
to include the _F_ flag hint.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Spreading the interrupts across the CPU cores is good for load
balancing, but not necessarily as good when using a CPU/core
that is not part of the NUMA local CPU. If it can be localized,
the kernel's cpumask_local_spread() service will pick a core
that is on the node close to the PCI device.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Replace the misguided "#pragma packed" with tags on each
struct/union definition that actually needs it. This is safer
and more efficient on the various compilers and architectures.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If the basic ionic interface works but the lif creation fails,
don't fail the probe. This will allow us to use the driver to
help inspect the hw/fw/pci interface for debugging purposes.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Florian Westphal says:
====================
mptcp: don't auto-adjust rcvbuf size if locked
The mptcp receive buffer is auto-sized based on the subflow receive
buffer. Don't do this if userspace specfied a value via SO_RCVBUF
setsockopt.
Also update selftest program to provide a new option to set a fixed
size.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The mptcp rcvbuf size is adjusted according to the subflow rcvbuf size.
This should not be done if userspace did set a fixed value.
Fixes: 600911ff5f72bae ("mptcp: add rmem queue accounting")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
allows to run the tests with fixed receive buffer by passing
"-R <value>" to mptcp_connect.sh.
While at it, add a default 10 second poll timeout so the "-t"
becomes optional -- this makes mptcp_connect simpler to use
during manual testing.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for configuring port mirroring through the cls_matchall
classifier. We do a full ingress and/or egress capture towards a
capture port.
MT7530 supports one monitor port and multiple mirrored ports.
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.open-mesh.org/linux-merge
Simon Wunderlich says:
====================
This cleanup patchset includes the following patches:
- bump version strings, by Simon Wunderlich
- Avoid RCU list-traversal in spinlock, by Sven Eckelmann
- Replace zero-length array with flexible-array member,
by Gustavo A. R. Silva
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Simon Wunderlich says:
====================
Here is a batman-adv bugfix:
- Don't schedule OGM for disabled interface, by Sven Eckelmann
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Heiner Kallweit says:
====================
r8169: series with improvements to rtl_tx
This series includes few improvements to rtl_tx().
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Until ae84bc187337 ("r8169: don't use bit LastFrag in tx descriptor
after send") we used to access another bit in the descriptor, therefore
it seems the barrier was needed. Since this commit DescOwn is the
only bit we're interested in, so the barrier isn't needed any longer.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Simplify the parameters taken by rtl8169_unmap_tx_skb, this makes
usage of this function easier to read and understand.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
So far tx_skb->skb is the only member of the two structs that is not
reset. Make understanding the code easier by resetting both structs
completely in rtl8169_unmap_tx_skb.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Slightly improve the code by converting this while to a for loop.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
What the driver writes into MAC_MAXLEN_CFG does not actually represent
VLAN_ETH_FRAME_LEN but instead ETH_FRAME_LEN + ETH_FCS_LEN. Yes they are
numerically equal, but the difference is important, as the switch treats
VLAN-tagged traffic specially and knows to increase the maximum accepted
frame size automatically. So it is always wrong to account for VLAN in
the MAC_MAXLEN_CFG register.
Unconditionally increase the maximum allowed frame size for
double-tagged traffic. Accounting for the additional length does not
mean that the other VLAN membership checks aren't performed, so there's
no harm done.
Also, stop abusing the MTU name for configuring the MRU. There is no
support for configuring the MRU on an interface at the moment.
Fixes: a556c76adc05 ("net: mscc: Add initial Ocelot switch support")
Fixes: fa914e9c4d94 ("net: mscc: ocelot: create a helper for changing the port MTU")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Commit e18b353f102e ("ipvlan: add cond_resched_rcu() while
processing muticast backlog") added a cond_resched_rcu() in a loop
using rcu protection to iterate over slaves.
This is breaking rcu rules, so lets instead use cond_resched()
at a point we can reschedule
Fixes: e18b353f102e ("ipvlan: add cond_resched_rcu() while processing muticast backlog")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Julian Wiedmann says:
====================
s390/qeth: updates 2020-03-06
please apply the following patch series for qeth to netdev's net-next
tree.
Just a small update to take care of a regression wrt to IRQ handling in
net-next, reported by Qian Cai. The fix needs some qdio layer changes,
so you will find Vasily's Acked-by in that patch.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
After recent cleanups this is just a complicated wrapper around an u32*.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Once the call to qdio_establish() has completed, qdio is free to deliver
data IRQs to the device driver's IRQ poll handler.
For qeth (the only qdio driver that currently uses IRQ polling) this is
problematic, since the IRQs can arrive before its NAPI instance is
even registered. Calling napi_schedule() from qeth_qdio_start_poll()
then crashes in various nasty ways.
Until recently qeth checked for IFF_UP to drop such early interrupts,
but that's fragile as well since it doesn't enforce any ordering.
Fix this properly by bringing up the qdio device in IRQS_DISABLED mode,
and have the driver explicitly opt-in to receive data IRQs.
qeth does so from qeth_open(), which kick-starts a NAPI poll and then
calls qdio_start_irq() from qeth_poll().
Also add a matching qdio_stop_irq() in qeth_stop() to switch the qdio
dataplane back into a disabled state.
Fixes: 3d35dbe6224e ("s390/qeth: don't check for IFF_UP when scheduling napi")
CC: Qian Cai <cai@lca.pw>
Reported-by: Qian Cai <cai@lca.pw>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Acked-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In our production environment we have faced with problem that updating
classid in cgroup with heavy tasks cause long freeze of the file tables
in this tasks. By heavy tasks we understand tasks with many threads and
opened sockets (e.g. balancers). This freeze leads to an increase number
of client timeouts.
This patch implements following logic to fix this issue:
аfter iterating 1000 file descriptors file table lock will be released
thus providing a time gap for socket creation/deletion.
Now update is non atomic and socket may be skipped using calls:
dup2(oldfd, newfd);
close(oldfd);
But this case is not typical. Moreover before this patch skip is possible
too by hiding socket fd in unix socket buffer.
New sockets will be allocated with updated classid because cgroup state
is updated before start of the file descriptors iteration.
So in common cases this patch has no side effects.
Signed-off-by: Dmitry Yakunin <zeil@yandex-team.ru>
Reviewed-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
