SKBs can come with a prioriy. Currently a priority of 0..7 is
assumed. But this assumption is incorrect. To fix this any
priority of 0 or higher then 7 will be adjusted by calling
cfg80211_classify8021d
Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Franky Lin <franky.lin@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Signed-off-by: Hante Meuleman <hante.meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Only hardcode the LED behavior if the SROM doesn't provide any for all
LEDs of the card. This avoids instantiating LED triggers for unconnected
LEDs, while (hopefully) keeping things working for old cards with a
blank SROM.
Signed-off-by: Lucas Stach <dev@lynxeye.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
gcc-6 reports:
drivers/net/wireless/marvell/libertas_tf/main.c:30:19: error:
'lbtf_driver_version' defined but not used
with -Werror=unused-const-variable=.
Reported-by: Fengguang Wu <fengguang.wu@intel.com> [0-day test robot]
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
There are nine copies of the _rtl88ee_read_adapter_info() function,
and most but not all of them cause a build warning in some configurations:
rtl8192de/hw.c: In function '_rtl92de_read_adapter_info':
rtl8192de/hw.c:1767:12: error: 'hwinfo' may be used uninitialized in this function [-Werror=maybe-uninitialized]
rtl8723ae/hw.c: In function '_rtl8723e_read_adapter_info.constprop':
rtlwifi/rtl8723ae/hw.c:1654:12: error: 'hwinfo' may be used uninitialized in this function [-Werror=maybe-uninitialized]
The problem is that when rtlefuse->epromtype is something other than
EEPROM_BOOT_EFUSE, the rest of the function uses undefined data, resulting
in random behavior later.
Apparently, in some drivers, the problem was already found and fixed
but the fix did not make it into the others.
This picks one approach to deal with the problem and applies identical
code to all 9 files, to simplify the later consolidation of those.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
There are two firmware events we handle similarly in brcmfmac:
BRCMF_E_LINK and BRCMF_E_IF. The difference from firmware point of view
is that the first one means BSS remains present in the firmware. Trying
to (re)create it (e.g. when adding new virtual interface) will result in
an error.
Current code treats both events in a similar way. It removes Linux
interface for each of them. It works OK with e.g. BCM43602. Its firmware
generates both events for each interface. It means we get BRCMF_E_LINK
and remove interface. That is soon followed by BRCMF_E_IF which means
BSS was also removed in a firmware. The only downside of this is a
harmless error like:
[ 208.643180] brcmfmac: brcmf_fweh_call_event_handler: no interface object
Unfortunately BCM4366 firmware doesn't automatically remove BSS and so
it doesn't generate BRCMF_E_IF. In such case we incorrectly remove Linux
interface on BRCMF_E_LINK as BSS is still present in the firmware. It
results in an error when trying to re-create virtual interface, e.g.:
> iw phy phy1 interface add wlan1-1 type __ap
[ 3602.929199] brcmfmac: brcmf_ap_add_vif: timeout occurred
command failed: I/O error (-5)
With this patch we don't remove Linux interface while firmware keeps
BSS. Thanks to this we keep a consistent states of host driver and
device firmware.
Further improvement should be to mark BSS as disabled and remove
interface on BRCMF_E_LINK. Then we should add support for reusing
BSS-es.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Firmware for new chipsets is based on a new major version of code
internally maintained at Broadcom. E.g. brcmfmac4366b-pcie.bin (used for
BCM4366B1) is based on 10.10.69.3309 while brcmfmac43602-pcie.ap.bin was
based on 7.35.177.56.
Currently setting AP 5 GHz channel doesn't work reliably with BCM4366B1.
When setting e.g. 36 control channel with VHT80 (center channel 42)
firmware may randomly pick one of:
1) 52 control channel with 58 as center one
2) 100 control channel with 106 as center one
3) 116 control channel with 122 as center one
4) 149 control channel with 155 as center one
It seems new firmwares require setting AP mode (BRCMF_C_SET_AP) before
specifying a channel. Changing an order of firmware calls fixes the
problem. This requirement resulted in two separated "chanspec" calls,
one in AP code path and one in P2P path.
This fix was verified with BCM4366B1 and tested for regressions on
BCM43602.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
gcc-6 on x86 started warning about wl3501_get_encode when building
with -O2:
drivers/net/wireless/wl3501_cs.c: In function ‘wl3501_get_encode’:
drivers/net/wireless/wl3501_cs.c:1769:5: warning: ‘implemented’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1686:19: warning: ‘threshold’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1702:20: warning: ‘threshold’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1719:23: warning: ‘txpow’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1752:20: warning: ‘retry’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1806:25: warning: ‘pwr_state’ may be used uninitialized in this function
drivers/net/wireless/wl3501_cs.c:1383:24: warning: ‘value’ may be used uninitialized in this function
I could not figure out what exactly confuses gcc here, but splitting the
wl3501_get_mib_value function into two helps the compiler to figure out
that the variables are not actually used uninitialized, and makes it
slightly clearer to a human reader what the function actually does and
which parts of it are under the spinlock.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The Documentation/devicetree/bindings/net/wireless/marvell-sd8xxx.txt DT
binding document lists the possible compatible strings that a SDIO child
node can have, so the driver checks if the defined in the node matches.
But the error message when that's not the case is misleading, so change
for one that makes clear what the error really is. Also, returning a -1
as errno code is not correct since that's -EPERM. A -EINVAL seems to be
a more appropriate one.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The Documentation/devicetree/bindings/net/wireless/marvell-sd8xxx.txt DT
binding document say that the "interrupts" property in the child node is
optional. So the property being missed shouldn't be treated as an error.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The function can fail so the returned value should be checked
and the error propagated to the caller in case of a failure.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
It's better to have the device name prefixed in the error message.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Instead of duplicating part of the cleanups needed in case of an error
in .probe callback, have a single error path and use goto labels as is
common practice in the kernel.
This also has the nice side effect that the cleanup operations are made
in the inverse order of their counterparts, which was not the case for
the mwifiex_add_card() error path.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
There's only a check if mwifiex_add_card() returned a nonzero value, but
the actual error code is neither stored nor propagated to the caller. So
instead of always returning -1 (which is -EPERM and not a suitable errno
code in this case), propagate the value returned by mwifiex_add_card().
Patch also removes the assignment of sdio_disable_func() returned value
since it was overwritten anyways and what matters is to know the error
value returned by the first function that failed.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
If the sdio_enable_func() function fails on .probe, the -EIO errno code
is always returned but that could make more difficult to debug and find
the cause of why the function actually failed.
Since the driver/device core prints the value returned by .probe in its
error message propagate what was returned by sdio_enable_func() at fail.
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
SDIO is an auto enumerable bus so the SDIO devices are matched using the
sdio_device_id table and not using compatible strings from a OF id table.
However, commit ce4f6f0c35 ("mwifiex: add platform specific wakeup
interrupt support") allowed to match nodes defined as child of the SDIO
host controller in the probe function using a compatible string to setup
platform specific parameters in the DT.
The problem is that the OF parse function is always called regardless if
the SDIO dev has an OF node associated or not, and prints an error if it
is not found. So, on a platform that doesn't have a node for a SDIO dev,
the following misleading error message will be printed:
[ 12.480042] mwifiex_sdio mmc2:0001:1: sdio platform data not available
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
This is helpful for debugging. Without this all I was getting from "iw"
command on failed creating of P2P interface was:
> command failed: Too many open files in system (-23)
Signed-off-by: Rafal Milecki <zajec5@gmail.com>
[arend@broadcom.com: reduce error prints upon iface creation]
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
This patch ensures that 'locally_generated' parameter is correctly
passed to cfg80211_disconnected() API.
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The GFP_DMA flag is obviously misunderstood in the mwifiex driver. It's
meant for legacy ISA DMA memory mappings only -- the lower 16MB on x86.
That doesn't apply to PCIe or SDIO devices, I guess.
Remove the GFP_DMA flag to reduce the need to place the socket buffer
allocation into the low mem DMA area, which might already be in use by
other drivers.
This misuse was flagged by the PaX USERCOPY feature by chance, as it
detected the user copy operation from a DMA buffer in the recvfrom()
syscall path.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Tested-by: Dennis Wassenberg <dennis.wassenberg@secunet.com>
Cc: Amitkumar Karwar <akarwar@marvell.com>
Cc: Nishant Sarmukadam <nishants@marvell.com>
Cc: Xinming Hu <huxm@marvell.com>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: Brad Spengler <spender@grsecurity.net>
Cc: PaX Team <pageexec@freemail.hu>
Acked-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
This is important for brcmfmac as some of released firmwares (e.g.
brcmfmac4366b-pcie.bin) may pick different channel than requested. This
has been tested with BCM4366B1 in D-Link DIR-885L.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Our d11 code supports encoding/decoding channel info into/from chanspec
format used by firmware. Current implementation is quite misleading
because of the way "chnum" field is used.
When encoding channel info, "chnum" has to be filled by a caller with
*center* channel number. However when decoding chanspec the same field
is filled with a *control* channel number.
1) This can be confusing. It's expected for information to be the same
after encoding and decoding.
2) It doesn't allow accessing all info when decoding. Some functions may
need to know both channel numbers, e.g. cfg80211 callback getting
current channel.
Solve this by adding a separated field for control channel.
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Reviewed-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Use kmemdup when some other buffer is immediately copied into allocated
region. It replaces call to allocation followed by memcpy, by a single
call to kmemdup.
Signed-off-by: Muhammad Falak R Wani <falakreyaz@gmail.com>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
gcc-6 warns about code in il3945_hw_txq_ctx_free() being
somewhat ambiguous:
drivers/net/wireless/intel/iwlegacy/3945.c:1022:5: warning: suggest explicit braces to avoid ambiguous 'else' [-Wparentheses]
This adds a set of curly braces to avoid the warning.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The action 'check for winner' and 'download firmware' should be an
atomic action. This is true for btmrvl driver but not mwmfiex, which
cause firmware download to fail when the following senerio happens:
1) mwifiex check winner status: true
2) btmrvl check winner status: true, and start downloading firmware
3) mwfieix tries to download firmware, but failed because btmrvl is
already downloading.
This won't happen if 1) and 3) is an atomic action. This patch adds
sdio_claim/release_host call around those two actions to make sure it's
atomic.
Signed-off-by: Wei-Ning Huang <wnhuang@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Simplify assignment in wlc_phy_rxcal_gainctrl_nphy_rev5.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Simplify _rtl92ee_phy_path_adda_on.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Do not dereference card before checking against NULL value.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Do not return undefined value for transmission power
if the rate is invalid.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Variable adapter is incorrectly initialized.
Fixes: bf00dc22bc ("mwifiex: AMSDU Rx frame handling in AP mode")
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Acked-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
msg is dereferenced before checking against NULL, e.g.
when assigning pad_bytes.
Remove the superfluous check in function rsi_mgmt_pkt_to_core.
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The old implementation was overcomplicated and slightly bugged in some
corner cases.
Consider following state of BSS-es (limited to 6 for simplification):
drvr->iflist[0]: { bsscfgidx:0, ndev->name:wlan1, }
drvr->iflist[1]: (null)
drvr->iflist[2]: { bsscfgidx:2, ndev->name:wlan1-1, }
drvr->iflist[3]: { bsscfgidx:3, ndev->name:wlan1-2, }
drvr->iflist[4]: (null)
drvr->iflist[5]: (null)
In such case the next AP interface should bsscfgidx 4 (we don't use 1 as
it's reserved for P2P).
With old code the loop iterations were following:
[ifidx = 0] [bsscfgidx = 2] [highest = 2]
[ifidx = 1] [bsscfgidx = 2] [highest = 2] available = true
[ifidx = 2] [bsscfgidx = 2] [highest = 2] bsscfgidx = highest + 1
[ifidx = 3] [bsscfgidx = 3] [highest = 2] bsscfgidx = highest + 1
[ifidx = 4] [bsscfgidx = 3] [highest = 2] available = true
[ifidx = 5] [bsscfgidx = 3] [highest = 2] available = true
There were 2 obvious problems:
1) Having empty BSS at index 1 was resulting in available being always
set to true, even if we would run out of BSS-es.
2) Calculated bsscfgidx was invalid (3 instead of 4) resulting in driver
not being able to create the 4th AP interface.
New code is simpler, placed in file where it's really used, handles
running out of free BSS-es and allows using 4 interfaces at the same
time. It also looks for the first free BSS instead of one after the last
in use. It works well with current driver (which doesn't allow deleting
interfaces) and should be future proof (if we ever allow deleting).
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Check the current PHY operating mode (gmode) to see if we should
fall back from 6MB OFDM to 11MB CCK. For 5GHz operation this isn't
allowed.
Note, the fallback lookup is only done for RTS rates; normal fallback
rates are done via mac80211 and aren't affected by this change.
Signed-off-by: Adrian Chadd <adrian@freebsd.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
brcmf_sdiod_intr_unregister call that removes both func1 and
func2 interrupt handlers only called when brcmf_ops_sdio_remove
is called for func 1 (which is the 2nd call) but sdio is expecting
it to be removed at the end of each sdio_remove call.
This is causing 'rmmod bcmrfmac' on a 4356-sdio chip to complain
with:
WARNING: driver brcmfmac did not remove its interrupt handler!
The modification makes calling brcmf_sdiod_intr_unregister multiple
times harmless by clearing the variables that track if interrupt
handlers have been installed, and then calls it on every
brcmf_ops_sdio_remove call instead of just remove for func 1.
Signed-off-by: Christian Daudt <csd@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
When chip attach fails, brcmf_sdiod_intr_unregister is being called
but that is too early as sdiodev->settings has not been set yet
nor has brcmf_sdiod_intr_register been called.
Change to use oob_irq_requested + newly created sd_irq_requested
to decide on what to unregister at intr_unregister time.
Steps to reproduce problem:
- modprobe brcmfmac using buggy FW
- rmmod brcmfmac
- modprobe brcmfmac again.
If done with a buggy firmware, brcm_chip_attach will fail on the
2nd modprobe triggering the call to intr_unregister and the
kernel oops when attempting to de-reference sdiodev->settings->bus.sdio
which has not yet been set.
Signed-off-by: Christian Daudt <csd@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
The previous text was confusing, leading readers to think this
driver was a duplicate, and so didn't need to be enabled.
After the removal of the older staging driver, this is the only
driver in mainline for these devices.
Signed-off-by: Lauri Kasanen <cand@gmx.com>
Acked-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
size_t objects should be printed with %Z printf format.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit a91eb52abb ("qed: Revisit chain implementation") contains an
incorrect implementation for BE platforms, as device's regpairs containing
addresses are LE and they're not converted correctly when read back.
In addition, it raises a compilation warning for 32-bit platforms where
dma_addr_t is a 32-bit variable.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yuval Mintz says:
====================
qed: RocE & iSCSI infrastructure
We plan on sending 2 new protocol drivers in the imminent future -
both our RoCE [qedr] and iSCSI [qedi] drivers. As both submissions
would be rather massive and in order to avoid collisions between them,
the common infrastructure on the qed side was prepared as an independent
patch-series to be sent ahead of those 2 submissions.
This patch series introduces in QED 2 new 'ids' - one for iscsi and
one for roce. It then goes and adds logic required for configuring
said protocols in HW. Notice it *doesn't* actually add any client using
said ids, but rather only the infrastructure to allow their later usage.
What this patch doesn't contain is the slowpath protocol-configuration
toward the firmware. I.e., it contains register-setting logic, memory
allocations, etc., but not actual flow-related configuration specific
to the protocl. Those would be sent as part of the protocol driver
submissions.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
RoCE and iSCSI would require some added/changed hw configuration in order
to properly run; The biggest single change being the requirement of
allocating and mapping host memory for several HW blocks that aren't being
used by qede [SRC, QM, TM, etc.].
In addition, whereas qede is only using context memory for HW blocks, the
new protocol would also require task memories to be added.
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds in the ecore 2 new personalities in addition to
QED_PCI_ETH - QED_PCI_ISCSI and QED_PCI_ETH_ROCE.
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds the qed portion of the RoCE & iSCSI firmware HSI,
as well as adding several new common HSI files which would be required
by both qed and qed* protocols.
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RoCE driver is going to need a 32-bit chain [current chain implementation
for qed* currently supports only 16-bit producer/consumer chains].
This patch adds said support, as well as doing other slight tweaks and
modifications to qed's chain API.
Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no reason in this lock. At least for now.
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Reviewed-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the more common kernel logging style and reduce object size.
The logging message prefix changes from a mixture of
"RxRPC:" and "RXRPC:" to "af_rxrpc: ".
$ size net/rxrpc/built-in.o*
text data bss dec hex filename
64172 1972 8304 74448 122d0 net/rxrpc/built-in.o.new
67512 1972 8304 77788 12fdc net/rxrpc/built-in.o.old
Miscellanea:
o Consolidate the ASSERT macros to use a single pr_err call with
decimal and hexadecimal output and a stringified #OP argument
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added a condition to avoid vlan devices with same MAC registering
as VF.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Marcelo Ricardo Leitner says:
====================
sctp: Add GSO support
This patchset adds sctp GSO support.
Performance tests indicates that increases throughput by 10% if using
bigger chunk sizes, specially if bigger than MTU. For small chunks, it
doesn't help much if not using heavy firewall rules.
For small chunks it will probably be of more use once we get something
like MSG_MORE as David Laight had suggested.
overall changes:
v1->v2:
Added support for receiving GSO frames on SCTP stack, as requested by
Dave Miller.
v2->v3:
Consider sctphdr size in skb_gso_transport_seglen()
rebased due to 5c7cdf339a ("gso: Remove arbitrary checks for
unsupported GSO")
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This is useful for debugging packet sizes.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP has this pecualiarity that its packets cannot be just segmented to
(P)MTU. Its chunks must be contained in IP segments, padding respected.
So we can't just generate a big skb, set gso_size to the fragmentation
point and deliver it to IP layer.
This patch takes a different approach. SCTP will now build a skb as it
would be if it was received using GRO. That is, there will be a cover
skb with protocol headers and children ones containing the actual
segments, already segmented to a way that respects SCTP RFCs.
With that, we can tell skb_segment() to just split based on frag_list,
trusting its sizes are already in accordance.
This way SCTP can benefit from GSO and instead of passing several
packets through the stack, it can pass a single large packet.
v2:
- Added support for receiving GSO frames, as requested by Dave Miller.
- Clear skb->cb if packet is GSO (otherwise it's not used by SCTP)
- Added heuristics similar to what we have in TCP for not generating
single GSO packets that fills cwnd.
v3:
- consider sctphdr size in skb_gso_transport_seglen()
- rebased due to 5c7cdf339a ("gso: Remove arbitrary checks for
unsupported GSO")
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is a preparation for the GSO one. In order to successfully
handle GSO packets on rx path we must not call skb_linearize, otherwise
it defeats any gain GSO may have had.
This patch thus delays as much as possible the call to skb_linearize,
leaving it to sctp_inq_pop() moment. For that the sanity checks
performed now know how to deal with fragments.
One positive side-effect of this is that if the socket is backlogged it
will have the chance of doing it on backlog processing instead of
during softirq.
With this move, it's evident that a check for non-linearity in
sctp_inq_pop was ineffective and is now removed. Note that a similar
check is performed a bit below this one.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
