This patch adds the following.
1. Adds 'clk_sclk_hdmi27m' clock to represent the HDMI 27MHz clock.
2. Adds 'clk_vpllsrc; clock of type clksrc_clk to represent the
input clock for VPLL.
3. Adds 'clk_sclk_vpll' clock of type clksrc_clk to represent the
output of the MUX_VPLL mux.
4. Add clk_sclk_hdmi27m, clk_vpllsrc and clk_sclk_vpll to the list
of clocks to be registered.
5. Adds boot time print of 'clk_sclk_vpll' clock rate.
6. Adds 'clk_fout_vpll' clock to plat-s5p such that it is reusable
on other s5p platforms.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_p83 clock, which is the PCLK clock for PSYS domain, is of
type 'struct clk' whereas on S5PV210, this clock is suitable to be
of type clksrc_clk clock (since it has a clock divider). So this
patch replaces the 'struct clk' type clock to 'struct clksrc_clk'
type clock for the PCLK PSYS clock.
This patch modifies the following.
1. Removes definitions and usage of 'clk_p66' clock.
2. Adds 'clk_pclk_psys' clock which is of type 'struct clksrc_clk'.
3. Replaces all usage of clk_p66 with clk_pclk_psys clock.
4. Adds clk_pclk_psys into list of clocks to be registered.
5. Removes the sys_clks array since it is no longer required.
Also the registration of clocks in sys_clks is also removed.
6. Remove the 'GET_DIV' as it is no longer required.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_p83 clock, which is the PCLK clock for DSYS domain, is of
type 'struct clk' whereas on S5PV210, this clock is suitable to be
of type clksrc_clk clock (since it has a clock divider). So this
patch replaces the 'struct clk' type clock to 'struct clksrc_clk'
type clock for the PCLK DSYS clock.
This patch modifies the following.
1. Remove definitions and usage of 'clk_p83' clock.
2. Adds 'clk_pclk_dsys' clock which is of type 'struct clksrc_clk'.
3. Replace all usage of clk_p83 with clk_pclk_dsys clock.
4. Adds clk_pclk_dsys into list of clocks to be registered.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_h100 clock represents the IMEM clock for the MSYS domain.
This clock rate of this clock is always half of the hclk_msys clock.
There is an issue when getting the clock rate of the clk_h100 clock
(clock get_rate hclk_h100 always returns clock rate that is equal to
the hclk_msys clock rate).
This patch modifies the following.
1. Moves the definition of the clk_h100 clock into the 'init_clocks'
list with the appropriate parent, ctrlbit, enable and ops fields.
2. The name of the clock is changed from 'clk_h100' to 'hclk_imem'
to represent more clearly that is represents the IMEM clock in
the MSYS domain.
3. The function to get the clock rate of the hclk_imem clock is added.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_p100 clock, which is the PCLK clock for MSYS domain, is of
type 'struct clk' whereas on S5PV210, this clock is suitable to be
of type clksrc_clk clock (since it has a choice of clock source
and a pre-divider). So this patch replaces the 'struct clk' type
clock to 'struct clksrc_clk' type clock for the PCLK MSYS clock.
This patch modifies the following.
1. Remove definitions and usage of 'clk_p100' clock.
2. Adds 'clk_pclk_msys' clock which is of type 'struct clksrc_clk'.
3. Replace all usage of clk_p100 with clk_pclk_msys clock.
4. Adds clk_pclk_msys into list of clocks to be registered.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_h133 clock, which is the HCLK clock for PSYS domain, is of
type 'struct clk' whereas on S5PV210, this clock is suitable to be
of type clksrc_clk clock (since it has a choice of clock source
and a pre-divider). So this patch replaces the 'struct clk' type
clock to 'struct clksrc_clk' type clock for the HCLK PSYS clock.
This patch modifies the following.
1. Remove definitions and usage of 'clk_h133' clock.
2. Adds 'clk_hclk_psys' clock which is of type 'struct clksrc_clk'.
3. Replace all usage of clk_h133 with clk_hclk_psys clock.
4. Adds clk_hclk_psys into list of clocks to be registered.
5. Removes the clock rate calculation of hclk133 and replaces
it with code that derives the HCLK PSYS clock rate from
the clk_hclk_psys clock.
6. Modify printing of the system clock rates.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_h166 clock, which is the HCLK clock for DSYS domain, is of
type 'struct clk' whereas on S5PV210, this clock is suitable to be
of type clksrc_clk clock (since it has a choice of clock source
and a pre-divider). So this patch replaces the 'struct clk' type
clock to 'struct clksrc_clk' type clock for the HCLK DSYS clock.
This patch modifies the following.
1. Remove definitions and usage of 'clk_h166' clock.
2. Adds 'clk_sclk_a2m' clock which is one of possible parent clock
sources for the DSYS HCLK clock.
3. Adds 'clk_hclk_dsys' clock which is of type 'struct clksrc_clk'.
4. Replace all usage of clk_h166 with clk_hclk_dsys clock.
5. Adds clk_sclk_a2m and clk_hclk_dsys into list of clocks to
be registered.
6. Removes the clock rate calculation of hclk166 and replaces
it with code that derives the HCLK DSYS clock rate from
the clk_hclk_dsys clock.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The clk_h200 represents the HCLK for the MSYS domain. This clock
is of type 'struct clk' but on V210, it is more suitable to be of
type 'struct clksrc_clk' (since it is a divided version of the
armclk). The replacement clock is renamed as clk_hclk_msys to
indicate that it represents the HCLK for MSYS domain.
This patch modifies the following.
1. Removes the usage of the clk_h200 clock.
2. Adds the new clock 'clk_hclk_msys'.
3. Adds clk_hclk_msys to the list of sysclks to be registered.
4. Modifies the hclk_msys clock rate calculation procedure to
be based on the new clk_hclk_msys clock.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
This patch modifies the following.
1. Adds arm clock 'clk_armclk' of type clksrc_clk clock type.
2. Adds arm clock to the list of system clocks 'sysclks' for
registering it along with other system clocks.
3. Modifies the armclk clock rate calculation procedure to be
based on the new clk_armclk clock.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The assignment of clock rates for fout apll/mpll/epll is moved further
up in the s5pv210_setup_clocks function because the subsequent patches
require the clock rate of fout clocks to be setup.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
This patch modifies the following.
1. Registers the mout_apll clksrc_clk clock.
2. The mout_mpll and mout_epll clocks were registered as 'struct clk'
types and then their parents were setup using the s3c_set_clksrc
function. This patch reduces the two steps into one by registering
the mout_mpll and mout_epll clocks using the s3c_register_clksrc
function.
3. As per point 2 above, the init_parents array is no longer required.
So the mout clocks are now put together in a new array named 'sysclks'.
The sysclks array will list the system level clocks and more
clocks will be added to it in the subsequent patches.
4. The clks array is left empty because of the movement of mpll and epll
clocks into the sysclks array. It is not deleted since subsequent
patches will add clocks into this array.
Signed-off-by: Thomas Abraham <thomas.ab <at> samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim <at> samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
The system clock definitions are currently defined below the
peripheral clock definitions in the V210 clock code. For the V210
clock updates that follow this patch, it is required that the
system clock definitions such as the mout_apll and mout_mpll be
defined prior to the device clock definitions. This patch
re-arranges the system clock defintions for the clock updates that
follow this patch.
Signed-off-by: Thomas Abraham <thomas.ab@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Now that cache_ioctl_procfs() calls the bkl explicitly, we need to
include the relevant header as well.
This fixes the following build error:
net/sunrpc/cache.c: In function 'cache_ioctl_procfs':
net/sunrpc/cache.c:1355: error: implicit declaration of function 'lock_kernel'
net/sunrpc/cache.c:1359: error: implicit declaration of function 'unlock_kernel'
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
There are no more users of procfs that implement the ioctl
callback. Drop the bkl from this path and warn on any use
of this callback.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: John Kacur <jkacur@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Push down the bkl from procfs's ioctl main handler to its users.
Only three procfs users implement an ioctl (non unlocked) handler.
Turn them into unlocked_ioctl and push down the Devil inside.
v2: PDE(inode)->data doesn't need to be under bkl
v3: And don't forget to git-add the result
v4: Use wrappers to pushdown instead of an invasive and error prone
handlers surgery.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: John Kacur <jkacur@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
In fact it is now added to the hot key list when newt_form__new is used,
allowing us to remove the explicit assignment in all its users.
The visible change is that <- will exit the menu that pops up when -> is
pressed (and Enter when callchains are not being used).
Suggested-by: Ingo Molnar <mingo@elte.hu>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
register_security() became __init function.
So do verify() and security_fixup_ops().
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <jmorris@namei.org>
This patch adds pathname grouping support, which is useful for grouping
pathnames that cannot be represented using /\{dir\}/ pattern.
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <jmorris@namei.org>
The ACPI dependency moved to the TPM, where it belongs. Although
IMA per-se does not require access to the bios measurement log,
verifying the IMA boot aggregate does, which requires ACPI.
This patch prereq's 'TPM: ACPI/PNP dependency removal'
http://lkml.org/lkml/2010/5/4/378.
Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
Reported-by: Jean-Christophe Dubois <jcd@tribudubois.net>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Tested-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: James Morris <jmorris@namei.org>
This patch pushes the ACPI dependency into the device driver code
itself. Now, even without ACPI/PNP enabled, the device can be registered
using the TIS specified memory space. This will however result in the
lack of access to the BIOS event log, being the only implication of such
ACPI removal.
Signed-off-by: Rajiv Andrade <srajiv@linux.vnet.ibm.com>
Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com>
Signed-off-by: James Morris <jmorris@namei.org>
Use kstrdup when the goal of an allocation is copy a string into the
allocated region.
The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression from,to;
expression flag,E1,E2;
statement S;
@@
- to = kmalloc(strlen(from) + 1,flag);
+ to = kstrdup(from, flag);
... when != \(from = E1 \| to = E1 \)
if (to==NULL || ...) S
... when != \(from = E2 \| to = E2 \)
- strcpy(to, from);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Acked-by: Eric Paris <eparis@redhat.com>
Signed-off-by: James Morris <jmorris@namei.org>
1) Allow transition to NEED RESET state only from READY state for ISP82xx.
2) Avoid infinite ISP aborts when chip reset fails.
3) Code cleanup to remove some of the unused debug code.
Signed-off-by: Lalit Chandivade <lalit.chandivade@qlogic.com>
Signed-off-by: Santosh Vernekar <santosh.vernekar@qlogic.com>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The problem occurring is a hw-race where there's an outstanding read-flash
operation occurring while the chip is being reset (done via an sg_reset script).
After the chip is paused, the read-flash operation never completes and the
DPC thread, while trying to complete the reset, is never able to recover,
as the HW appears to be hung...
The fix is to wait for outstanding flash operation prior to doing a sg_reset -h.
And to wait for reset to complete before any flash operations.
Note, during the wait, if any of the operation (reset/flash) does not complete,
failure is returned to the upper layer. The upper layer either need to
fail or retry.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The qla2xxx driver uses a port_id_t to mark the start of its enumerations. gcc
is complaining that wrap.b24 may be used uninitialized, but this doesn't look
to be possible. Silence it.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Currently we can only issue the task management (TM)
commands via the mailbox mechanism. This is a limitation,
since only one mailbox command can be issued at a time.
The purpose of this effort is to provide support for
issuing and processing the respose to TM and Marker
IOCBs asynchronously. Towards achieving this, the
consolidated srb architecture that is currently used for
BSG and IOCB/Logio commands has been enhanced and used.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Currently, BSG and IOCB/Logio commands have a different
framework (srb structs). The purpose of this effort is to
consolidate them into a generalized framework for these
as well as other asynchronous operations in the future.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Migrate to a consistent set of debug entries during status-IOCB
handling:
* group CS_TIMEOUT handling with CS_PORT_UNAVAILABLE and the like
(more regrouping of common behaviour).
* drop CS_DATA_OVERRUN handling as it now falls into the
'default' case (still returns DID_ERROR).
* consolidate CS_RESET and CS_ABORTED handling, as we the only
functional difference was a printk() (still returns DID_RESET).
* dropped all the earlier inconsistent [qla_]printk()s sprinkled
throught the needlessly large case-statement. Failure case
I/Os are now logged with a 'standard' format:
<command failure details>
<command generic details>
so, for example a dropped-frame is logged as:
qla2xxx 0000:13:00.0: scsi(16:0:0) Dropped frame(s) detected (4000 of 4000 bytes).
qla2xxx 0000:13:00.0: scsi(16:0:0) FCP command status: 0x15-0x18 (70018) \
oxid=78 ser=76 cdb=280000 len=4000 rsp_info=0 resid=0 fw_resid=4000
which should now convey all relevant information.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The driver should not be unloaded if any application is using it.
To disallow driver unload, driver use count must be incremented.
Application uses this char device as handle and increases driver
use count to avoid possible driver unload.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Currently error status is cleared only after the uncorrectable fatal errors
in the qla2xxx_pci_slot_reset. This fix is added to clear the error status in
qla2xxx_pci_resume. This way for both fatal and non-fatal errors the error
status gets cleared properly.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=572258
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
With qla2xxx using mid-layer async-scsi-scanning, the link state for
disconnected port is displayed wrong. Additional check for cable presence
is considered to display proper link state.
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
This update handles a use case where pm_qos update requests need to
silently fail if the update is being sent to a handle that is NULL.
The problem was that the original pm_qos silently fails when a request
update is passed to a parameter that has not been added to the list yet.
This update restores that behavior.
Signed-off-by: markgross <markgross@thegnar.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
commit 70b25f890c
Author: Tejun Heo <tj@kernel.org>
Date: Thu Apr 15 09:00:08 2010 +0900
[SCSI] fix locking around blk_abort_request()
Introduced a reference before check problem, fix this by moving the
lock shorthand code to be right at the point of actual use.
Reported-by: Dan Carpenter <error27@gmail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
rtnetlink: make SR-IOV VF interface symmetric
sctp: delete active ICMP proto unreachable timer when free transport
tcp: fix MD5 (RFC2385) support
* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
MIPS: Oprofile: Fix Loongson irq handler
MIPS: N32: Use compat version for sys_ppoll.
MIPS FPU emulator: allow Cause bits of FCSR to be writeable by ctc1
Now we have a set of nested attributes:
IFLA_VFINFO_LIST (NESTED)
IFLA_VF_INFO (NESTED)
IFLA_VF_MAC
IFLA_VF_VLAN
IFLA_VF_TX_RATE
This allows a single set to operate on multiple attributes if desired.
Among other things, it means a dump can be replayed to set state.
The current interface has yet to be released, so this seems like
something to consider for 2.6.34.
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Synchronize access to the drivers configuration interface.
Also do not allow configuration changes during online/offline
transition.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
z/OS may activate Optimized Latency Mode (OLM) for a connection
through an OSA Express3 adapter, which reduces the number of
allowed concurrent connections, if adapter is used in shared mode.
Create a meaningful message, if activation of an OSA-connection fails
due to an active OLM-connection on the shared OSA-adapter.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
OSA supports HW TX checksumming in layer 3 mode. Enable this
feature and remove software fallback used for TSO. Cleanup
checksum bits to indicate OSA can do checksumming only for
IPv4 TCP and UDP.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
transport may be free before ICMP proto unreachable timer expire, so
we should delete active ICMP proto unreachable timer when transport
is going away.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
vlan/macvlan start_xmit() can inform caller of congestion with
NET_XMIT_CN return value. This doesnt mean packet was dropped.
Increment normal stat counters instead of tx_dropped.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP-MD5 sessions have intermittent failures, when route cache is
invalidated. ip_queue_xmit() has to find a new route, calls
sk_setup_caps(sk, &rt->u.dst), destroying the
sk->sk_route_caps &= ~NETIF_F_GSO_MASK
that MD5 desperately try to make all over its way (from
tcp_transmit_skb() for example)
So we send few bad packets, and everything is fine when
tcp_transmit_skb() is called again for this socket.
Since ip_queue_xmit() is at a lower level than TCP-MD5, I chose to use a
socket field, sk_route_nocaps, containing bits to mask on sk_route_caps.
Reported-by: Bhaskar Dutta <bhaskie@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP MD5 support uses percpu data for temporary storage. It currently
disables preemption so that same storage cannot be reclaimed by another
thread on same cpu.
We also have to make sure a softirq handler wont try to use also same
context. Various bug reports demonstrated corruptions.
Fix is to disable preemption and BH.
Reported-by: Bhaskar Dutta <bhaskie@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With RPS inclusion, skb timestamping is not consistent in RX path.
If netif_receive_skb() is used, its deferred after RPS dispatch.
If netif_rx() is used, its done before RPS dispatch.
This can give strange tcpdump timestamps results.
I think timestamping should be done as soon as possible in the receive
path, to get meaningful values (ie timestamps taken at the time packet
was delivered by NIC driver to our stack), even if NAPI already can
defer timestamping a bit (RPS can help to reduce the gap)
Tom Herbert prefer to sample timestamps after RPS dispatch. In case
sampling is expensive (HPET/acpi_pm on x86), this makes sense.
Let admins switch from one mode to another, using a new
sysctl, /proc/sys/net/core/netdev_tstamp_prequeue
Its default value (1), means timestamps are taken as soon as possible,
before backlog queueing, giving accurate timestamps.
Setting a 0 value permits to sample timestamps when processing backlog,
after RPS dispatch, to lower the load of the pre-RPS cpu.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
