| Age | Commit message (Collapse) | Author |
|---|
|
This change is meant ot prevent stealing the skb->head to use as a page in
the event that the skb->head was cloned. This allows the other clones to
track each other via shinfo->dataref.
Without this we break down to two methods for tracking the reference count,
one being dataref, the other being the page count. As a result it becomes
difficult to track how many references there are to skb->head.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Extend tcp coalescing implementing it from tcp_queue_rcv(), the main
receiver function when application is not blocked in recvmsg().
Function tcp_queue_rcv() is moved a bit to allow its call from
tcp_data_queue()
This gives good results especially if GRO could not kick, and if skb
head is a fragment.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Before stealing fragments or skb head, we must make sure skbs are not
cloned.
Alexander was worried about destination skb being cloned : In bridge
setups, a driver could be fooled if skb->data_len would not match skb
nr_frags.
If source skb is cloned, we must take references on pages instead.
Bug happened using tcpdump (if not using mmap())
Introduce kfree_skb_partial() helper to cleanup code.
Reported-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
An EEH error can cause the FW to trigger a flash debug dump.
Resetting the card while flash dump is in progress can cause it not to recover.
Wait for it to finish before letting EEH flow to reset the card.
Signed-off-by: Sathya Perla <Sathya.Perla@emulex.com>
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: Sarveshwar Bandi <Sarveshwar.Bandi@emulex.com>
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Suggested-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Sarveshwar Bandi <sarveshwar.bandi@emulex.com>
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This renders the interface view somewhat inconsistent from the Host OS POV
considering the rest of the interfaces are showing their respective speeds
based on the bandwidth assigned to them.
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implementing the advanced early retransmit (sysctl_tcp_early_retrans==2).
Delays the fast retransmit by an interval of RTT/4. We borrow the
RTO timer to implement the delay. If we receive another ACK or send
a new packet, the timer is cancelled and restored to original RTO
value offset by time elapsed. When the delayed-ER timer fires,
we enter fast recovery and perform fast retransmit.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch implements RFC 5827 early retransmit (ER) for TCP.
It reduces DUPACK threshold (dupthresh) if outstanding packets are
less than 4 to recover losses by fast recovery instead of timeout.
While the algorithm is simple, small but frequent network reordering
makes this feature dangerous: the connection repeatedly enter
false recovery and degrade performance. Therefore we implement
a mitigation suggested in the appendix of the RFC that delays
entering fast recovery by a small interval, i.e., RTT/4. Currently
ER is conservative and is disabled for the rest of the connection
after the first reordering event. A large scale web server
experiment on the performance impact of ER is summarized in
section 6 of the paper "Proportional Rate Reduction for TCP”,
IMC 2011. http://conferences.sigcomm.org/imc/2011/docs/p155.pdf
Note that Linux has a similar feature called THIN_DUPACK. The
differences are THIN_DUPACK do not mitigate reorderings and is only
used after slow start. Currently ER is disabled if THIN_DUPACK is
enabled. I would be happy to merge THIN_DUPACK feature with ER if
people think it's a good idea.
ER is enabled by sysctl_tcp_early_retrans:
0: Disables ER
1: Reduce dupthresh to packets_out - 1 when outstanding packets < 4.
2: (Default) reduce dupthresh like mode 1. In addition, delay
entering fast recovery by RTT/4.
Note: mode 2 is implemented in the third part of this patch series.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This a prepartion patch that refactors the code to enter recovery
into a new function tcp_enter_recovery(). It's needed to implement
the delayed fast retransmit in ER.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fix this compiler warning (on PowerPC) by not marking a parameter as
const:
drivers/net/ethernet/pasemi/pasemi_mac.c: In function 'pasemi_mac_replenish_rx_ring':
drivers/net/ethernet/pasemi/pasemi_mac.c:646:3: warning: passing argument 1 of 'netdev_alloc_skb' discards qualifiers from pointer target type
include/linux/skbuff.h:1706:31: note: expected 'struct net_device *' but argument is of type 'const struct net_device *'
Cc: Olof Johansson <olof@lixom.net>
Cc: Pradeep A. Dalvi <netdev@pradeepdalvi.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
commit 30a5de7723a8a4211be02e94236e9167a424fd07 added
ability to use single MSI-X vector, but lack proper
handling for 57710/57711 HW
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If the user request for the number of VFs in the max_vfs parameter is
out of range then reset the value to the default value of zero. This
makes the behavior of the ixgbe driver the same as for the igb driver.
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Robert Garrett <robertx.e.garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
If the host VMM administrator has set the virtual function device's
MAC address then also deny VF requests for MACVLAN filters.
Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Garrett, Robert <robertx.e.garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Some of our adapters have thermal data available, this patch exports
this data via hwmon sysfs interface.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Some 82599 adapters contain thermal data that we can get to via
an i2c interface. These functions provide support to get at that
data. A following patch will export this data.
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Secondary unicast and multicast addresses are added to the Receive
Address registers (RAR) for most parts supported by the driver. For
82579, there is only one actual RAR and a number of Shared Receive Address
registers (SHRAR) that are shared among the driver and f/w which can be
reserved and write-protected by the f/w. On this device, use the SHRARs
that are not taken by f/w for the additional addresses.
Add a MAC ops function pointer infrastructure (similar to other MAC
operations in the driver) for setting RARs, introduce a new rar_set
function for 82579 and convert the existing code that sets RARs on other
devices to a generic rar_set function.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
The PHY initialization flows and assorted workarounds for 82577/8/9 done
during driver load and resume from Sx should be the same yet they are not.
Combine the current flows/workarounds into a common set of functions that
are called during the different code paths.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
An update to the EEPROM on 82579 will extend a delay in hardware to fix an
issue with WoL not working after a G3->S5 transition which is unrelated to
the driver. However, this extended delay conflicts with nominal operation
of the device when it is initialized by the driver and after every reset
of the hardware (i.e. the driver starts configuring the device before the
hardware is done with it's own configuration work). The workaround for
when the driver is in control of the device is to tell the hardware after
every reset the configuration delay should be the original shorter one.
Some pre-existing variables are renamed generically to be re-used with
new register accesses.
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
|
|
Add ECN (Explicit Congestion Notification) marking capability to netem
tc qdisc add dev eth0 root netem drop 0.5 ecn
Instead of dropping packets, try to ECN mark them.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Hagen Paul Pfeifer <hagen@jauu.net>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Hagen Paul Pfeifer <hagen@jauu.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
remove useless casts and rename variables for less confusion.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
TCP or UDP stacks have big enough latencies that prefetching next
pointer is worth it.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The netlink API lets users create unmanaged L2TPv3 tunnels using
iproute2. Until now, a request to create an unmanaged L2TPv3 IP
encapsulation tunnel over IPv6 would be rejected with
EPROTONOSUPPORT. Now that l2tp_ip6 implements sockets for L2TP IP
encapsulation over IPv6, we can add support for that tunnel type.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
L2TPv3 defines an IP encapsulation packet format where data is carried
directly over IP (no UDP). The kernel already has support for L2TP IP
encapsulation over IPv4 (l2tp_ip). This patch introduces support for
L2TP IP encapsulation over IPv6.
The implementation is derived from ipv6/raw and ipv4/l2tp_ip.
Signed-off-by: Chris Elston <celston@katalix.com>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
For implementing other protocols on top of IPv6, such as L2TPv3's IP
encapsulation over ipv6, we'd like to call some IPv6 functions which
are not currently exported. This patch exports them.
Signed-off-by: Chris Elston <celston@katalix.com>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds support for unmanaged L2TPv3 tunnels over IPv6 using
the netlink API. We already support unmanaged L2TPv3 tunnels over
IPv4. A patch to iproute2 to make use of this feature will be
submitted separately.
Signed-off-by: Chris Elston <celston@katalix.com>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If an L2TP tunnel uses IPv6, make sure the l2tp debugfs file shows the
IPv6 address correctly.
Signed-off-by: Chris Elston <celston@katalix.com>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Userspace uses connect() to associate a pppol2tp socket with a tunnel
socket. This needs to allow the caller to supply the new IPv6
sockaddr_pppol2tp structures if IPv6 is used.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Checkpatch warns about the use of __attribute__((packed)). So use the
recommended __packed syntax instead.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The l2tp_ip socket currently maintains packet/byte stats in its
private socket structure. But these counters aren't exposed to
userspace and so serve no purpose. The counters were also
smp-unsafe. So this patch just gets rid of the stats.
While here, change a couple of internal __u32 variables to u32.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Cleanup the l2tp_ip code to make use of an existing ipv4 support function.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
L2TP uses 64-bit counters but since these are not updated atomically,
we need to make them safe for smp. This patch addresses that.
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
PHY polling code for FPGA is considered in every MDIO R/W API.
no need to add additional code to atl1c_change_mtu.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: David Liu <dwliu@qca.qaulcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
L0S might be unstable if no cable link, only enable it when link up.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There may be tx-skbs still pending in HW when PHY link down.
Reset MAC will make the DMA engine go to the start point.
and release all pending skbs.
Note: Reset MAC will clear any interrupt status and mask.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
common_task might be running while close routine is called,
wait/cancel it.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The hardware incorrectly process L0S/L1 entrance if the chipset/root
response after specific/shorter timer and cause system hang.
Enlarge the timeout value to avoid this issue.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
On some platform with EEPROM/OTP existing, the BIOS could overwrite
a new MAC address for the NIC. so, the permanent mac address should
be from BIOS. the address is restored when driver removing.
Voltage raising isn't applicable for l1d.
Replace swab32 with htonl for big/little endian platform.
related Registers are refined as well.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The Close-action is done by atl1c_reset_pcie, remove it from
atl1c_get_permanent_address.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
WoL status is read-clear and should be cleared when in S0
status.
putting it in atl1c_reset_pcie is more suitable than
in atl1c_get_permanent_address.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
On some platforms the PHY settings need to change depending on the
cable link status to get better stability.
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Tested-by: Liu David <dwliu@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
All supported devices have one issue that msi interrupt doesn't assert
if pci command register bit (PCI_COMMAND_INTX_DISABLE) is set.
Add workaround in drivers/pci/quirks.c
Signed-off-by: xiong <xiong@qca.qualcomm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux
|
|
Before doing skb->head_frag work on bnx2x driver, I found too much stuff
was inlined in bnx2x/bnx2x_cmn.h for no good reason and made my work not
very easy.
Move some big functions out of this include file to the respective .c
file.
A lot of inline keywords are not needed at all in this huge driver.
text data bss dec hex filename
490083 1270 56 491409 77f91 bnx2x/bnx2x.ko.before
484206 1270 56 485532 7689c bnx2x/bnx2x.ko
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The reset logic after a Rx FIFO overrun will clear the programmed
multicast addresses. This patch fixes the issue by reprogramming the
registers after the reset.
The commit eefc48b ("pch_gbe: reprogram multicast address register on
reset") tried to fix this problem, but it introduces unnecessary
codes. In fact, all multicast addresses have been saved in netdev->mc,
So we can call pch_gbe_set_multi() directly after reset_hw and
reset_rx.
This commit kills 50+ line codes
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Takahiro Shimizu <tshimizu818@gmail.com>
Signed-off-by: RongQing.Li <roy.qing.li@gmail.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
__skb_splice_bits() can check if skb to be spliced has its skb->head
mapped to a page fragment, instead of a kmalloc() area.
If so we can avoid a copy of the skb head and get a reference on
underlying page.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
TCP coalesce can check if skb to be merged has its skb->head mapped to a
page fragment, instead of a kmalloc() area.
We had to disable coalescing in this case, for performance reasons.
We 'upgrade' skb->head as a fragment in itself.
This reduces number of cache misses when user makes its copies, since a
less sk_buff are fetched.
This makes receive and ofo queues shorter and thus reduce cache line
misses in TCP stack.
This is a followup of patch "net: allow skb->head to be a page fragment"
Tested with tg3 nic, with GRO on or off. We can see "TCPRcvCoalesce"
counter being incremented.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
GRO can check if skb to be merged has its skb->head mapped to a page
fragment, instead of a kmalloc() area.
We 'upgrade' skb->head as a fragment in itself
This avoids the frag_list fallback, and permits to build true GRO skb
(one sk_buff and up to 16 fragments), using less memory.
This reduces number of cache misses when user makes its copy, since a
single sk_buff is fetched.
This is a followup of patch "net: allow skb->head to be a page fragment"
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch converts tg3 driver, one of our reference drivers, to use new
build_skb() api in frag mode.
Instead of using kmalloc() to allocate the memory block that will be
used by build_skb() as skb->head, we use a page fragment.
This is a followup of patch "net: allow skb->head to be a page fragment"
This allows GRO, TCP coalescing, and splice() to be more efficient.
Incidentally, this also removes SLUB slow path contention in kfree()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
skb->head is currently allocated from kmalloc(). This is convenient but
has the drawback the data cannot be converted to a page fragment if
needed.
We have three spots were it hurts :
1) GRO aggregation
When a linear skb must be appended to another skb, GRO uses the
frag_list fallback, very inefficient since we keep all struct sk_buff
around. So drivers enabling GRO but delivering linear skbs to network
stack aren't enabling full GRO power.
2) splice(socket -> pipe).
We must copy the linear part to a page fragment.
This kind of defeats splice() purpose (zero copy claim)
3) TCP coalescing.
Recently introduced, this permits to group several contiguous segments
into a single skb. This shortens queue lengths and save kernel memory,
and greatly reduce probabilities of TCP collapses. This coalescing
doesnt work on linear skbs (or we would need to copy data, this would be
too slow)
Given all these issues, the following patch introduces the possibility
of having skb->head be a fragment in itself. We use a new skb flag,
skb->head_frag to carry this information.
build_skb() is changed to accept a frag_size argument. Drivers willing
to provide a page fragment instead of kmalloc() data will set a non zero
value, set to the fragment size.
Then, on situations we need to convert the skb head to a frag in itself,
we can check if skb->head_frag is set and avoid the copies or various
fallbacks we have.
This means drivers currently using frags could be updated to avoid the
current skb->head allocation and reduce their memory footprint (aka skb
truesize). (thats 512 or 1024 bytes saved per skb). This also makes
bpf/netfilter faster since the 'first frag' will be part of skb linear
part, no need to copy data.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Matt Carlson <mcarlson@broadcom.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
