sta32x resets and loses all configuration during ESD test.
Work around by polling the CONFA register once a second
and restore all coeffcients and registers when CONFA
changes unexpectedly.
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Add a structure for platform specific configuration and use it,
thereby removing a few FIXMEs which marked hard-coded values.
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The coefficient RAM must be saved in a shadow so it can
be restored when the codec is powered on using
regulator_bulk_enable().
Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: stable@vger.kernel.org
It takes a while to find the macro-magically defined Page*()
functions defined in include/linux/page-flags.h if you're new to
the kernel. Add some magic to the tags script to transform these
macros into the actual functions they are, so that tag jumping in
the mm code is a bit easier.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
commit de47725 (include: replace linux/module.h
with "struct module" wherever possible) introduced
a compilation breaked when it removed <linux/module.h>
from <linux/device.h> which musb glue layers were
(mistakenly) relying on.
Include that header to fix the compile error.
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Without this the gadget will never be able to allocate a stream capable
endpoint. The manual says that the stream id is a 16bit id. It does not
talk about an upper limit in any other way. So I think 15 is a
reasonable limit :)
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
new endpoint should start from DATA0,
but mod_host didn't care it.
This patch fix it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
renesas_usbhs is caring pipe type and its direction.
but current usbhs_endpoint_alloc() didn't check direction.
this patch modify it.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since commit "193ab2a usb: gadget: allow multiple gadgets to be built"
the udc controllers can be compiled as a module. The ci13xxx_msm driver
is missing the MODULE_LICENSE statement, so loading fails with:
ci13xxx_msm: module license 'unspecified' taints kernel.
ci13xxx_msm: Unknown symbol dev_set_name (err 0)
ci13xxx_msm: Unknown symbol platform_get_irq (err 0)
ci13xxx_msm: Unknown symbol device_unregister (err 0)
ci13xxx_msm: Unknown symbol usb_add_gadget_udc (err 0)
ci13xxx_msm: Unknown symbol put_device (err 0)
ci13xxx_msm: Unknown symbol platform_driver_register (err 0)
ci13xxx_msm: Unknown symbol platform_get_resource (err 0)
ci13xxx_msm: Unknown symbol device_register (err 0)
ci13xxx_msm: Unknown symbol usb_del_gadget_udc (err 0)
This patch adds the missing MODULE_LICENSE statement with GPL v2 according
to the header of the driver.
Tested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch fixes the following compile errors that show up if switching
on the debug trace code:
drivers/usb/gadget/ci13xxx_udc.c: In function 'ci13xxx_wakeup':
drivers/usb/gadget/ci13xxx_udc.c:2517:3: error: 'dev' undeclared (first use in this function)
drivers/usb/gadget/ci13xxx_udc.c:2517:3: note: each undeclared identifier is reported only once for each function it appears in
drivers/usb/gadget/ci13xxx_udc.c: In function 'udc_probe':
drivers/usb/gadget/ci13xxx_udc.c:2867:2: error: 'name' undeclared (first use in this function)
Tested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The inline documentation of _gadget_stop_activity() states that the
function should be called holding the udc->lock. This however will
result in a deadlock, because _gadget_stop_activity() takes the udc->lock.
During normal operation _gadget_stop_activity() is always called unlocked,
but in ci13xxx_stop() it's called locked, this results in the following
deadlock during rmmod of a gadget driver.
This patch fixes the deadlock by calling _gadget_stop_activity() always
unlocked, the inline documentation is adjusted accordingly.
=============================================
[ INFO: possible recursive locking detected ]
3.1.0-rc6+ #159
---------------------------------------------
rmmod/121 is trying to acquire lock:
(udc_lock){-.-...}, at: [<c0229048>] _gadget_stop_activity+0x18/0x154
but task is already holding lock:
(udc_lock){-.-...}, at: [<c02291e0>] ci13xxx_stop+0x5c/0x164
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(udc_lock);
lock(udc_lock);
*** DEADLOCK ***
May be due to missing lock nesting notation
2 locks held by rmmod/121:
#0: (udc_lock#2){+.+.+.}, at: [<c02286c0>] usb_gadget_unregister_driver+0x34/0x88
#1: (udc_lock){-.-...}, at: [<c02291e0>] ci13xxx_stop+0x5c/0x164
stack backtrace:
[<c000d41c>] (unwind_backtrace+0x0/0xf0) from [<c0056f94>] (check_deadlock.clone.24+0x284/0x2c4)
[<c0056f94>] (check_deadlock.clone.24+0x284/0x2c4) from [<c00589ac>] (validate_chain.clone.25+0x430/0x6fc)
[<c00589ac>] (validate_chain.clone.25+0x430/0x6fc) from [<c0059bac>] (__lock_acquire+0x494/0x8f0)
[<c0059bac>] (__lock_acquire+0x494/0x8f0) from [<c005a698>] (lock_acquire+0x98/0x1a8)
[<c005a698>] (lock_acquire+0x98/0x1a8) from [<c02f12a4>] (_raw_spin_lock_irqsave+0x64/0xa0)
[<c02f12a4>] (_raw_spin_lock_irqsave+0x64/0xa0) from [<c0229048>] (_gadget_stop_activity+0x18/0x154)
[<c0229048>] (_gadget_stop_activity+0x18/0x154) from [<c0229234>] (ci13xxx_stop+0xb0/0x164)
[<c0229234>] (ci13xxx_stop+0xb0/0x164) from [<c022867c>] (usb_gadget_remove_driver+0x88/0x98)
[<c022867c>] (usb_gadget_remove_driver+0x88/0x98) from [<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88)
[<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88) from [<c0065f2c>] (sys_delete_module+0x1fc/0x26c)
[<c0065f2c>] (sys_delete_module+0x1fc/0x26c) from [<c00092a0>] (ret_fast_syscall+0x0/0x38)
BUG: spinlock lockup on CPU#0, rmmod/121, c05b1644
[<c000d41c>] (unwind_backtrace+0x0/0xf0) from [<c01da000>] (do_raw_spin_lock+0x128/0x144)
[<c01da000>] (do_raw_spin_lock+0x128/0x144) from [<c02f12c8>] (_raw_spin_lock_irqsave+0x88/0xa0)
[<c02f12c8>] (_raw_spin_lock_irqsave+0x88/0xa0) from [<c0229048>] (_gadget_stop_activity+0x18/0x154)
[<c0229048>] (_gadget_stop_activity+0x18/0x154) from [<c0229234>] (ci13xxx_stop+0xb0/0x164)
[<c0229234>] (ci13xxx_stop+0xb0/0x164) from [<c022867c>] (usb_gadget_remove_driver+0x88/0x98)
[<c022867c>] (usb_gadget_remove_driver+0x88/0x98) from [<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88)
[<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88) from [<c0065f2c>] (sys_delete_module+0x1fc/0x26c)
[<c0065f2c>] (sys_delete_module+0x1fc/0x26c) from [<c00092a0>] (ret_fast_syscall+0x0/0x38)
Tested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The current driver sets the request's dma addr (mReq->req.dma) to 0 to
mark the DMA address as not valid. However some gadget drivers
(e.g. gadgetfs) set the request's dma addr to DMA_ADDR_INVALID to mark
the address as invalid. This leads to bogus data send because the
ci13xxx_udc driver assumes the request has already been mapped.
This patch fixes the problem, by using DMA_ADDR_INVALID instead of 0
to mark the request's DMA address as invalid.
Tested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Under certain circumstances lockdep finds an inconsistent lock state in
gadgetfs. The problem can be reproduced with a hardware using the
ci13xxx_udc driver and the gadgetfs test program (needs a patch to support
the ci13xxx_udc, though):
http://www.linux-usb.org/gadget/usb.c
Start the test program, wait to initialize, then press Ctrl+c.
This patch fixes the following problem by using spin_lock_irqsave()
instead of spin_lock().
=================================
[ INFO: inconsistent lock state ]
3.1.0-rc6+ #158
---------------------------------
inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
usb/113 [HC0[0]:SC0[0]:HE1:SE1] takes:
(&(&dev->lock)->rlock){?.....}, at: [<bf000340>] gadgetfs_disconnect+0x14/0x80 [gadgetfs]
{IN-HARDIRQ-W} state was registered at:
[<c00596b8>] mark_irqflags+0x14c/0x1ac
[<c0059bf8>] __lock_acquire+0x4e0/0x8f0
[<c005a698>] lock_acquire+0x98/0x1a8
[<c02f10e0>] _raw_spin_lock+0x54/0x8c
[<bf000340>] gadgetfs_disconnect+0x14/0x80 [gadgetfs]
[<c0229104>] _gadget_stop_activity+0xd4/0x154
[<c022b130>] isr_reset_handler+0x34/0x1c0
[<c022c320>] udc_irq+0x204/0x228
[<c0069018>] handle_irq_event_percpu+0x64/0x3a0
[<c0069390>] handle_irq_event+0x3c/0x5c
[<c006ae5c>] handle_level_irq+0x8c/0x10c
[<c0068a34>] generic_handle_irq+0x30/0x44
[<c0009b2c>] handle_IRQ+0x30/0x84
[<c0008ef8>] __irq_svc+0x38/0x60
[<c0009c58>] default_idle+0x30/0x34
[<c0009e30>] cpu_idle+0x9c/0xd8
[<c04056f4>] start_kernel+0x278/0x2bc
irq event stamp: 6412
hardirqs last enabled at (6412): [<c02f1cd0>] _raw_spin_unlock_irqrestore+0x30/0x5c
hardirqs last disabled at (6411): [<c02f1278>] _raw_spin_lock_irqsave+0x20/0xa0
softirqs last enabled at (6381): [<c002833c>] irq_exit+0xa0/0xa8
softirqs last disabled at (6372): [<c002833c>] irq_exit+0xa0/0xa8
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(&dev->lock)->rlock);
<Interrupt>
lock(&(&dev->lock)->rlock);
*** DEADLOCK ***
1 lock held by usb/113:
#0: (udc_lock#2){+.+.+.}, at: [<c02286c0>] usb_gadget_unregister_driver+0x34/0x88
stack backtrace:
[<c000d41c>] (unwind_backtrace+0x0/0xf0) from [<c0058e98>] (print_usage_bug+0x144/0x1c4)
[<c0058e98>] (print_usage_bug+0x144/0x1c4) from [<c0059144>] (mark_lock_irq+0x22c/0x274)
[<c0059144>] (mark_lock_irq+0x22c/0x274) from [<c00592d4>] (mark_lock+0x148/0x3e0)
[<c00592d4>] (mark_lock+0x148/0x3e0) from [<c0059668>] (mark_irqflags+0xfc/0x1ac)
[<c0059668>] (mark_irqflags+0xfc/0x1ac) from [<c0059bf8>] (__lock_acquire+0x4e0/0x8f0)
[<c0059bf8>] (__lock_acquire+0x4e0/0x8f0) from [<c005a698>] (lock_acquire+0x98/0x1a8)
[<c005a698>] (lock_acquire+0x98/0x1a8) from [<c02f10e0>] (_raw_spin_lock+0x54/0x8c)
[<c02f10e0>] (_raw_spin_lock+0x54/0x8c) from [<bf000340>] (gadgetfs_disconnect+0x14/0x80 [gadgetfs])
[<bf000340>] (gadgetfs_disconnect+0x14/0x80 [gadgetfs]) from [<c0229104>] (_gadget_stop_activity+0xd4/0x154)
[<c0229104>] (_gadget_stop_activity+0xd4/0x154) from [<c0229240>] (ci13xxx_stop+0xbc/0x17c)
[<c0229240>] (ci13xxx_stop+0xbc/0x17c) from [<c022867c>] (usb_gadget_remove_driver+0x88/0x98)
[<c022867c>] (usb_gadget_remove_driver+0x88/0x98) from [<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88)
[<c02286f4>] (usb_gadget_unregister_driver+0x68/0x88) from [<bf0003e8>] (dev_release+0x14/0x48 [gadgetfs])
[<bf0003e8>] (dev_release+0x14/0x48 [gadgetfs]) from [<c00cc158>] (__fput+0xa4/0x1f0)
[<c00cc158>] (__fput+0xa4/0x1f0) from [<c00c7f28>] (filp_close+0x5c/0x74)
[<c00c7f28>] (filp_close+0x5c/0x74) from [<c00c7fe8>] (sys_close+0xa8/0x150)
[<c00c7fe8>] (sys_close+0xa8/0x150) from [<c00092a0>] (ret_fast_syscall+0x0/0x38)
Tested-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The dynamic jack allocation of the MIDI gadget currently links all
external jacks to one single instance of an embedded jack. According to
the spec, this is only valid if these streams always carry the same data
stream, as described in the USB MIDI 1.0 spec, chapter 3.3.1.
Also, genius Windows 7(tm) terminates it's life cycle instantly with a
blue screen of death once a device with more than one input and output
port with the current implementation is connected.
While at it, and because it grew again by this change, allocate the
temporary function pointer list on the heap, not on the stack.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The USB-IF CV compliance tester is getting stricter, and it would
be valid for it to fail a mass-storage device that accepts an
invalid USB_BULK_RESET_REQUEST request. Although it doesn't do
that yet, let's be proactive and fix that now.
Suggested by Alan Stern.
Signed-off-by: Paul Zimmerman <paulz@synopsys.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The latest USB-IF CV tester checks for a valid length for this
request.
Signed-off-by: Paul Zimmerman <paulz@synopsys.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix up the following section mismatch warnings:
WARNING: drivers/usb/renesas_usbhs/renesas_usbhs.o(.text+0xf5d): Section
mismatch in reference from the function usbhs_mod_probe() to the function
.devinit.text:usbhs_mod_host_probe() The function usbhs_mod_probe() references
the function __devinit usbhs_mod_host_probe(). This is often because
usbhs_mod_probe lacks a __devinit annotation or the annotation of
usbhs_mod_host_probe is wrong.
WARNING: drivers/usb/renesas_usbhs/renesas_usbhs.o(.text+0xfd7): Section
mismatch in reference from the function usbhs_mod_probe() to the function
.devexit.text:usbhs_mod_host_remove() The function usbhs_mod_probe() references
a function in an exit section. Often the function usbhs_mod_host_remove() has
valid usage outside the exit section and the fix is to remove the __devexit
annotation of usbhs_mod_host_remove.
WARNING: drivers/usb/renesas_usbhs/renesas_usbhs.o(.text+0x1005): Section
mismatch in reference from the function usbhs_mod_remove() to the function
.devexit.text:usbhs_mod_host_remove() The function usbhs_mod_remove()
references a function in an exit section. Often the function
usbhs_mod_host_remove() has valid usage outside the exit section and the fix is
to remove the __devexit annotation of usbhs_mod_host_remove.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since renesas_usbhs mod_host didn't use
struct completion as static object, the warning of lockdep came out.
This patch fixup this issue.
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix compile error in file drivers/usb/gadget/fsl_udc_core.c.
drivers/usb/gadget/fsl_udc_core.c: In function 'portscx_device_speed':
drivers/usb/gadget/fsl_udc_core.c:1720: error: 'speed' undeclared (first
use in this function)
Introduced in commit e538dfdae8
(usb: Provide usb_speed_string() function)
Signed-off-by: Alexander Aring <a.aring@phytec.de>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The "BCLR" in CFIFOCTR/DnFIFOCTR can flush the fifo of "CPU side" only.
To flush the fifo of "SIE side", we have to use the "ACLRM" in PIPEnCTR.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The udc-newstyle needs device_register in probe() of platform_device.
If it doesn't call, kernel panic happens in the sysfs_create_dir() when
we run modprobe a gadget driver.
[ balbi@ti.com : fix compile warning introduced by this patch ]
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Those are simply giving information about the current
state of the UDC, nothing really fancy. We can let
everybody read those.
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
usb_gadget_disconnect() is responsible of removing
data pullups. Before doing that we must, first, tell
gadget driver we're disconnecting (by calling disconnect
method on gadget driver structure), unbind the gadget
driver and stop the controller.
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
that has already being done by udc-core.c. It's
unnecessary and might cause issues with some gadget
drivers.
Tested: Ajay Kumar Gupta <ajay.gupta@ti.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We should not be using dev_get_drvdata() because we
never call dev_set_drvdata(). Let's use container_of()
as all other sysfs attributes.
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
when dev_hard_header() failed, the newly allocated skb should be freed.
Signed-off-by: RongQing.Li <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The support for NLM_F_* flags at IPv6 routing requests.
If NLM_F_CREATE flag is not defined for RTM_NEWROUTE request,
warning is printed, but no error is returned. Instead new route is
added. Later NLM_F_CREATE may be required for
new route creation.
Exception is when NLM_F_REPLACE flag is given without NLM_F_CREATE, and
no matching route is found. In this case it should be safe to assume
that the request issuer is familiar with NLM_F_* flags, and does really
not want route to be created.
Specifying NLM_F_REPLACE flag will now make the kernel to search for
matching route, and replace it with new one. If no route is found and
NLM_F_CREATE is specified as well, then new route is created.
Also, specifying NLM_F_EXCL will yield returning of error if matching
route is found.
Patch created against linux-3.2-rc1
Signed-off-by: Matti Vaittinen <Mazziesaccount@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The support for NLM_F_* flags at IPv6 routing requests.
Warn if NLM_F_CREATE flag is not defined for RTM_NEWROUTE request,
creating new table. Later NLM_F_CREATE may be required for
new route creation.
Patch created against linux-3.2-rc1
Signed-off-by: Matti Vaittinen <Mazziesaccount@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Lantiq ETOP ethernet driver fails to build in 3.2-rc1 due to 2 missing
header files.
Signed-off-by: John Crispin <blogic@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
* 'for-usb-linus' of ra.kernel.org:/pub/scm/linux/kernel/git/sarah/xhci:
usb, xhci: Clear warm reset change event during init
xhci: Set slot and ep0 flags for address command.
usb, xhci: fix lockdep warning on endpoint timeout
This patch fixes an issue introduced recently with commit
452448f928.
CC: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When HW doesn't remove FCS bytes they are reported in the completion
byte count, we don't need to take them to skb.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
This round of floor sweeping converts strncpy calls in various .get_drvinfo
routines to the preferred strlcpy. It also does a modicum of other
cleaning in those routines.
Signed-off-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bnx2x uses following formula to compute its rx_buf_sz :
dev->mtu + 2*L1_CACHE_BYTES + 14 + 8 + 8 + 2
Then core network adds NET_SKB_PAD and SKB_DATA_ALIGN(sizeof(struct
skb_shared_info))
Final allocated size for skb head on x86_64 (L1_CACHE_BYTES = 64,
MTU=1500) : 2112 bytes : SLUB/SLAB round this to 4096 bytes.
Since skb truesize is then bigger than SK_MEM_QUANTUM, we have lot of
false sharing because of mem_reclaim in UDP stack.
One possible way to half truesize is to reduce the need by 64 bytes
(2112 -> 2048 bytes)
Instead of allocating a full cache line at the end of packet for
alignment, we can use the fact that skb_shared_info sits at the end of
skb->head, and we can use this room, if we convert bnx2x to new
build_skb() infrastructure.
skb_shared_info will be initialized after hardware finished its
transfert, so we can eventually overwrite the final padding.
Using build_skb() also reduces cache line misses in the driver, since we
use cache hot skb instead of cold ones. Number of in-flight sk_buff
structures is lower, they are recycled while still hot.
Performance results :
(820.000 pps on a rx UDP monothread benchmark, instead of 720.000 pps)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Eilon Greenstein <eilong@broadcom.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
CC: Tom Herbert <therbert@google.com>
CC: Jamal Hadi Salim <hadi@mojatatu.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
CC: Thomas Graf <tgraf@infradead.org>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
One of the thing we discussed during netdev 2011 conference was the idea
to change some network drivers to allocate/populate their skb at RX
completion time, right before feeding the skb to network stack.
In old days, we allocated skbs when populating the RX ring.
This means bringing into cpu cache sk_buff and skb_shared_info cache
lines (since we clear/initialize them), then 'queue' skb->data to NIC.
By the time NIC fills a frame in skb->data buffer and host can process
it, cpu probably threw away the cache lines from its caches, because lot
of things happened between the allocation and final use.
So the deal would be to allocate only the data buffer for the NIC to
populate its RX ring buffer. And use build_skb() at RX completion to
attach a data buffer (now filled with an ethernet frame) to a new skb,
initialize the skb_shared_info portion, and give the hot skb to network
stack.
build_skb() is the function to allocate an skb, caller providing the
data buffer that should be attached to it. Drivers are expected to call
skb_reserve() right after build_skb() to adjust skb->data to the
Ethernet frame (usually skipping NET_SKB_PAD and NET_IP_ALIGN, but some
drivers might add a hardware provided alignment)
Data provided to build_skb() MUST have been allocated by a prior
kmalloc() call, with enough room to add SKB_DATA_ALIGN(sizeof(struct
skb_shared_info)) bytes at the end of the data without corrupting
incoming frame.
data = kmalloc(NET_SKB_PAD + NET_IP_ALIGN + 1536 +
SKB_DATA_ALIGN(sizeof(struct skb_shared_info)),
GFP_ATOMIC);
...
skb = build_skb(data);
if (!skb) {
recycle_data(data);
} else {
skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
...
}
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Eilon Greenstein <eilong@broadcom.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
CC: Tom Herbert <therbert@google.com>
CC: Jamal Hadi Salim <hadi@mojatatu.com>
CC: Stephen Hemminger <shemminger@vyatta.com>
CC: Thomas Graf <tgraf@infradead.org>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The "#include <linux/module.h>" was replaced by "#include <linux/export.h>"
in the patch "powerpc: various straight conversions from module.h --> export.h".
This will cause the following compile problem:
arch/powerpc/sysdev/fsl_rio.c: In function 'fsl_rio_mcheck_exception':
arch/powerpc/sysdev/fsl_rio.c:296: error: implicit declaration of function 'search_exception_tables'.
The file fsl_rio.c needs the declaration of function "search_exception_tables"
in the header file "linux/module.h".
Signed-off-by: Liu Gang <Gang.Liu@freescale.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Previously we claimed device ID 0x7450, regardless of the vendor, which is
clearly wrong. Now we'll claim that device ID only for AMD.
I suspect this was just a typo in the original code, but it's possible this
change will break shpchp on non-7450 AMD bridges. If so, we'll have to fix
them as we find them.
Reference: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=638863
Reported-by: Ralf Jung <ralfjung-e@gmx.de>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The Windows driver .inf disables ASPM on hpsa devices. Do the same because the
selection of a non default ASPM policy can cause the device to hang.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: stable@kernel.org
Acked-by: Mike Miller <mike.miller@hp.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
The ALC882 macpro and imac24 static configs can be transferred to the
auto-parser with the additional GPIO setup.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In the commit [c3e837bb: ALSA: hda/realtek - Rewrite ALC882 acer-aspire-*
models with the auto-parser], the check of the model option got removed
mistakenly. Re-added the board_config check again.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When the pin or the DAC doesn't support the stereo, create a mono
control instead of creating a stereo control blindly.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The code located at arch/mips/lantiq/ included module.h to be able to use
the EXPORT_SYMBOL* macros. These can now be directly included using
export.h.
Signed-off-by: John Crispin <blogic@openwrt.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/2937/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
