This GPIO controller is used on UniPhier SoC family.
The vendor specific property "socionext,interrupt-ranges" is for
specifying interrupt mapping to the parent interrupt controller
because the mapping is not contiguous. It works like "ranges",
but transforms "interrupts" instead of "reg".
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The legacy block layer handles requests as follows:
- If the prep function returns BLKPREP_OK, let blk_peek_request()
return the pointer to that request.
- If the prep function returns BLKPREP_DEFER, keep the RQF_STARTED
flag and retry calling the prep function later.
- If the prep function returns BLKPREP_KILL or BLKPREP_INVALID, end
the request.
In none of these cases it is correct to clear the SCMD_INITIALIZED
flag from inside scsi_prep_fn(). Since scsi_prep_fn() already
guarantees that scsi_init_command() will be called once even if
scsi_prep_fn() is called multiple times, remove the code that clears
SCMD_INITIALIZED from scsi_prep_fn().
The scsi-mq code handles requests as follows:
- If scsi_mq_prep_fn() returns BLKPREP_OK, set the RQF_DONTPREP flag
and submit the request to the SCSI LLD.
- If scsi_mq_prep_fn() returns BLKPREP_DEFER, call
blk_mq_delay_run_hw_queue() and return BLK_STS_RESOURCE.
- If the prep function returns BLKPREP_KILL or BLKPREP_INVALID, call
scsi_mq_uninit_cmd() and let the blk-mq core end the request.
In none of these cases scsi_mq_prep_fn() should clear the
SCMD_INITIALIZED flag. Hence remove the code from scsi_mq_prep_fn()
function that clears that flag.
This patch avoids that the following warning is triggered when using
the legacy block layer:
------------[ cut here ]------------
WARNING: CPU: 1 PID: 4198 at drivers/scsi/scsi_lib.c:654 scsi_end_request+0x1de/0x220
CPU: 1 PID: 4198 Comm: mkfs.f2fs Not tainted 4.14.0-rc5+ #1
task: ffff91c147a4b800 task.stack: ffffb282c37b8000
RIP: 0010:scsi_end_request+0x1de/0x220
Call Trace:
<IRQ>
scsi_io_completion+0x204/0x5e0
scsi_finish_command+0xce/0xe0
scsi_softirq_done+0x126/0x130
blk_done_softirq+0x6e/0x80
__do_softirq+0xcf/0x2a8
irq_exit+0xab/0xb0
do_IRQ+0x7b/0xc0
common_interrupt+0x90/0x90
</IRQ>
RIP: 0010:_raw_spin_unlock_irqrestore+0x9/0x10
__test_set_page_writeback+0xc7/0x2c0
__block_write_full_page+0x158/0x3b0
block_write_full_page+0xc4/0xd0
blkdev_writepage+0x13/0x20
__writepage+0x12/0x40
write_cache_pages+0x204/0x500
generic_writepages+0x48/0x70
blkdev_writepages+0x9/0x10
do_writepages+0x34/0xc0
__filemap_fdatawrite_range+0x6c/0x90
file_write_and_wait_range+0x31/0x90
blkdev_fsync+0x16/0x40
vfs_fsync_range+0x44/0xa0
do_fsync+0x38/0x60
SyS_fsync+0xb/0x10
entry_SYSCALL_64_fastpath+0x13/0x94
---[ end trace 86e8ef85a4a6c1d1 ]---
Fixes: commit 64104f7032 ("scsi: Call scsi_initialize_rq() for filesystem requests")
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Damien Le Moal <damien.lemoal@wdc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Correct spelling error.
Signed-off-by: Kevin Barnett <kevin.barnett@microsemi.com>
Signed-off-by: Don Brace <don.brace@microsemi.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add JEDEC entry for the Winbond w25q16fw/w25q16dw with similar
flags and format than the Winbond w25q32dw entry.
Tested on a Khadas VIM2 SBC board with an Amlogic S912 SoC.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Cyrille Pitchen <cyrille.pitchen@wedev4u.fr>
Remove unnecessary braces for single statements also declare a local
variable "platform_state" of type "struct vchiq_2835_state", to
avoid the multiple cast.
Signed-off-by: Mihaela Muraru <mihaela.muraru21@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ever since it became possible to compile the SCSI core code as a module,
the documentation describing the module parameters has been incorrect.
Update the documentation to add a "scsi_mod." prefix to the relevant
options.
Reported-by: Laurence Oberman <loberman@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
An independent security researcher, Mohamed Ghannam, has reported
this vulnerability to Beyond Security's SecuriTeam Secure Disclosure
program.
The xfrm_dump_policy_done function expects xfrm_dump_policy to
have been called at least once or it will crash. This can be
triggered if a dump fails because the target socket's receive
buffer is full.
This patch fixes it by using the cb->start mechanism to ensure that
the initialisation is always done regardless of the buffer situation.
Fixes: 12a169e7d8 ("ipsec: Put dumpers on the dump list")
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
This patch introduces a new ap_qact() function which
exploits the PQAP(QACT) subfunction. QACT is a new
interface to Query the Ap Compatilibity Type based
on a given AP qid, type, mode and version.
Based on this new function the AP bus scan code is
slightly reworked to use this new interface for
querying the compatible type for each new AP queue
device detected. So new and unknown devices can
get automatically mapped to a compatible type and
handled without the need for toleration patches
for every new hardware.
The currently highest known hardware is CEX6S.
With this patch a possible successor can get
queried for a combatible type known by the device
driver without the need for an toleration patch.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
With the CEX6 there is a new CPRB (subfunction AU) used
to generate protected keys from secure keys. This new
CPRB needs to have the special flag set in the queue
message header struct which is introduced with this fix.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch adds the full CEX6S card support to the zcrypt device
driver. A CEX6A/C/P is detected and displayed as such, the card
and queue device driver code is updated to recognize it and the
relative weight values for CEX4, CEX5 and CEX6 have been updated.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch changes the way the primary surface is used for dumb
framebuffers. Instead of configuring the bo itself as primary surface
a shadow bo is created and used instead. Framebuffers can share the
shadow bo in case they have the same format and resolution.
On atomic plane updates we don't have to update the primary surface in
case we pageflip from one framebuffer to another framebuffer which
shares the same shadow. This in turn avoids the flicker caused by the
primary-destroy + primary-create cycle, which is very annonying when
running wayland on qxl.
The qxl driver never actually writes to the shadow bo. It sends qxl
blit commands which update it though, and the spice server might
actually execute them (and thereby write to the shadow) in case the
local rendering is kicked for some reason. This happens for example in
case qemu is asked to write out a dump of the guest display (screendump
monitor command).
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20171019062150.28090-3-kraxel@redhat.com
To avoid introduction of a new compatible for each small SoC/DMA controller
variation, move the definition of the channel count to the devicetree.
The number of vchans is no longer explicit, but limited by the highest
port/DMA request number. The result is a slight overallocation for SoCs
with a sparse port mapping.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
By default, we build Broadcom SBA RAID driver as loadable module for
iProc SOCs so that kernel image is little smaller and we load SBA RAID
driver only when required.
Signed-off-by: Anup Patel <anup.patel@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This patch makes the comment header of Broadcom SBA RAID driver
similar to the GPL comment header used across Broadcom driver
sources.
Signed-off-by: Anup Patel <anup.patel@broadcom.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Each mailbox channel used by Broadcom SBA RAID driver is
a separate HW ring.
Currently, Broadcom SBA RAID driver creates one DMA channel
using one or more mailbox channels. When we are using more
than one mailbox channels for a DMA channel, the sba_request
are distributed evenly among multiple mailbox channels which
results in sba_request being completed out-of-order.
The above described out-of-order completion of sba_request
breaks the dma_async_is_complete() API because it assumes
DMA cookies are completed in orderly fashion.
To ensure correct behaviour of dma_async_is_complete() API,
this patch updates Broadcom SBA RAID driver to use only
single mailbox channel. If additional mailbox channels are
specified in DT then those will be ignored.
Signed-off-by: Anup Patel <anup.patel@broadcom.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
As-per documentation in driver/dma/dmaengine.h, the
dma_cookie_complete() API should be called with lock
held.
This patch ensures that Broadcom SBA RAID driver calls
the dma_cookie_complete() API with reqs_lock held.
Signed-off-by: Anup Patel <anup.patel@broadcom.com>
Reviewed-by: Ray Jui <ray.jui@broadcom.com>
Reviewed-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Some quirky UDCs (like dwc3 on Exynos) need to have their phys calibrated e.g.
for using super speed. This patch adds a new phy_calibrate() method.
When the calibration should be used is dependent on actual chip.
In case of dwc3 on Exynos the calibration must happen after usb_add_hcd()
(while in host mode), because certain phy parameters like Tx LOS levels
and boost levels need to be calibrated further post initialization of xHCI
controller, to get SuperSpeed operations working. But an hcd must be
prepared first in order to pass it to usb_add_hcd(), so, in particular, dwc3
registers must be available first, and in order for the latter to happen
the phys must be initialized. This poses a chicken and egg problem if
the calibration were to be performed in phy_init(). To break the circular
dependency a separate method is added which can be called at a desired
moment after phy intialization.
Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Refactor ufs_qcom_power_up_sequence() to get rid of ugly
exported phy APIs and use the phy_init() and phy_power_on()
to do the phy initialization.
Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Set the phy mode based on the UFS HS PA mode. This lets the
controller let phy know the mode in which the PHY Adapter is
running and set the phy rates accordingly.
Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Reviewed-by: Subhash Jadavani <subhashj@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Adding support to set desired UFS phy mode that can be set
from the host controller.
Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
UFS phy has two modes for each High speed generation.
These modes are identified by two rates of operations -
Rate A, and Rate B.
Add these UFS phy modes to phy framework.
Signed-off-by: Vivek Gautam <vivek.gautam@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Update the PCIe phy settings based on new settings available
in AM572x Technical Reference Manual[1] Revision I, revised
April 2017 in Table 26-62 "Preferred PCIe_PHY_RX SCP Register
Settings".
[1] http://www.ti.com/lit/ug/spruhz6i/spruhz6i.pdf
Cc: Vignesh R <vigneshr@ti.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
[nsekhar@ti.com: commit message updates]
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
PCI core access configuration space registers in resume_noirq callbacks.
In the case of dra7xx, PIPE3 PHY connected to PCIe controller has to be
enabled before accessing configuration space registers. Since
PIPE3 PHY is enabled by only configuring control module registers, no
aborts has been observed so far (though during noirq stage, interface
clock of PIPE3 PHY is not enabled).
With new TRM updates, PIPE3 PHY has to be initialized (PIPE3 PHY
registers has to be accessed) as well which requires the interface
clock of PIPE3 PHY to be enabled. The interface clock of PIPE3 PHY is
derived from OCP2SCP and hence PCIe PHY is modeled as a child of
OCP2SCP. Since pm_runtime is not enabled during noirq stage,
pm_runtime_get_sync done in phy_init doesn't enable
OCP2SCP clocks resulting in abort when PIPE3 PHY registers are
accessed.
Create a function dependency between PCIe and PHY here to make
sure PCIe is suspended before PCIe PHY/OCP2SCP and resumed after
PCIe PHY/OCP2SCP.
Suggested-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Parse the DT properties brcm,rxaeq-mode and brcm,rxaeq-value to
correctly configure the RX equalizer of the PHY. This may be required to
resolve specific signal integrity issues.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Split the functional code in brcm_stb_sata_init() to a separate function
that actually does configure spread spectrum: brcm_stb_sata_ssc_init()
and make that function return void, since that function cannot fail.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Define two new properties: brcm,rx-aeq-mode which allows configuring the
SATA PHY RX equalizers and when "manual" is used, brcm,rx-aeq can be
used to set the exact value.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
This patch adds binding for r8a77995 (R-Car D3). Since r8a77995 doesn't
have dedicated pins (ID, VBUS), this will match against the generic
fallback on R-Car D3.
For now, this driver doesn't support usb role swap for r8a77995.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
This patch adds SoC-specific parameter to avoid reading/writing
specific registers wrongly if this driver runs on a SoC which doesn't
have dedicated pins (e.g. R-Car D3). This patch also changes the
value "has_otg" to "has_otg_pins" for slightly easier reading of
the code.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
This patch modifies the role_store() to use "enum phy_mode" instead
of the local "bool" for host/device mode selection.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
The previous code assumed a channel has otg capability if a channel
has interrupt property. But, it is not good because:
- Battery charging feature also needs interrupt property.
- Some R-Car Gen3 SoCs (e.g. R-Car D3) don't have OTG capability.
So, this patch checks whether usb 2.0 host node has dr_mode property or
not. If it has 'dr_mode = "otg";', this driver enables otg capability.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Add USB PHY support for r8a7743/5 SoC. Renesas RZ/G1[ME] (R8A7743/5)
USB PHY is identical to the R-Car Gen2 family.
Signed-off-by: Biju Das <biju.das@bp.renesas.com>
Acked-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Rob Herring <robh@kernel.org>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Allwinner V3s SoC also features the dual route of the first USB PHY.
Enable it.
Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Calculate the calibration code as per the docs. The docs talk about
reading and averaging the pullup and pulldown calibration codes. They
also talk about adding in some adjustment codes. Let's do what the
docs say.
In practice this doesn't seem to matter a whole lot. On a device I
tested the pullup and pulldown codes were nearly the same (0x23 and
0x24) and the adjustment codes were 0.
Reviewed-by: Chris Zhong <zyw@rock-chips.com>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
NOTE: nothing is known to be fixed by this change, but it does enforce
some delays that are documented to be necessary. Possibly this could
fix some corner cases.
The function tcphy_dp_aux_calibration(), like most of the functions in
the type C PHY, is mostly undocumented and filled with mysterious,
hardcoded numbers.
Let's attempt to try to document some of these numbers and clean the
function up a little bit. Here's the actual cleanup that happened
here:
1. All magic numbers were replaced with bit definitions.
2. For registers that we modify multiple times I now keep track of the
value of the register rather than randomly doing a
read/modify/write or just hardcoding a new number based on knowing
what the old number was.
3. Delay 10 ms (vs 1 ms) after writing the calibration code. No idea
if this is important but it matches the example in the docs.
4. Whenever setting a "delayed" version of a signal always put an
explicit delay in the code. No known problems were seen without
this delay but it seems wise to have it. Whenever a delay of "at
least 100 ns" was specified I used a delay of 1 us.
5. Added comments to some of the bits of code.
6. Removed duplicate setting of TX_ANA_CTRL_REG_5 (to 0)
7. Moved setting of TX_ANA_CTRL_REG_3 to the same place it was in the
sample code. Note that TX_ANA_CTRL_REG_3 ought to be initted to 0
(and elsewhere we assume that we just got a reset), but it seems
fine to be explicit.
8. Treats the calibration code as a 7-bit two's complement number.
This isn't strictly required, but seems slightly cleaner. The docs
say "treat this as a two's complement number, but it should never
be negative". If we ever read the "adjustment" codes as documented
then perhaps the two's complement bit will matter more.
There are still a few weird / mysterious things around aux init and
this doesn't attempt to fix all of them. Mostly it's aimed at doing
changes that should be _very_ safe and add a lot of clarity. Things
specifically not done:
A) Resolve the fact that some registers are read/modify/write and
others are explicitly initted to a value. We always call
tcphy_dp_aux_calibration() right after resetting the PHY so it's
probably not critical, but it's a little weird that the code is
inconsistent.
B) Fully resolve the documented init sequence with the current one.
We still have a few mystery steps and we also leave out turning on
TXDA_DRV_LDO_BG_FB_EN and TXDA_DRV_LDO_BG_REF_EN, which is in the
sample code.
C) Clean things up to read all the bits of the calibration code. This
will hopefully come in a followup change.
This also doesn't attempt to document any of the other parts of the
PHY--just the aux init which is all I got docs for.
Reviewed-by: Chris Zhong <zyw@rock-chips.com>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
If we're trying to get a handle to an optional phy, then the phy framework
being disabled shouldn't return an hard error.
Instead, return NULL just like phy_optional_get does when there's no phy
provided in the DT.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
This is used to force PHY with USB OTG function to enter a specific
mode, and override OTG IDPIN(or IDDIG) signal.
Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
The 'modes' member of the mvebu_comphy_priv structure is not used.
Remove it.
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
The variable "ret" is immediately reassigned by a following statement.
Thus omit the explicit initialisation at the beginning.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
The array extcon_info is read only, local to the source and does not
need to be in global scope, so make it static const.
Cleans up sparse warning:
symbol 'extcon_info' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
The extcon has two type of extcon devices as following.
- 'extcon provider deivce' adds new extcon device and detect the
state/properties of external connector. Also, it notifies the
state/properties to the extcon consumer device.
- 'extcon consumer device' gets the change state/properties
from extcon provider device.
Prior to that, include/linux/extcon.h contains all exported API for
both provider and consumer device driver. To clarify the meaning of
header file and to remove the wrong use-case on consumer device,
this patch separates into extcon.h and extcon-provider.h.
[Description for include/linux/{extcon.h|extcon-provider.h}]
- extcon.h includes the extcon API and data structure for extcon consumer
device driver. This header file contains the following APIs:
: Register/unregister the notifier to catch the change of extcon device
: Get the extcon device instance
: Get the extcon device name
: Get the state of each external connector
: Get the property value of each external connector
: Get the property capability of each external connector
- extcon-provider.h includes the extcon API and data structure for extcon
provider device driver. This header file contains the following APIs:
: Include 'include/linux/extcon.h'
: Allocate the memory for extcon device instance
: Register/unregister extcon device
: Set the state of each external connector
: Set the property value of each external connector
: Set the property capability of each external connector
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Acked-by: Sebastian Reichel <sebastian.reichel@collabora.co.uk>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Acked-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Acked-by: Lee Jones <lee.jones@linaro.org>
Acked-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Acked-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
This patch fixes the following lockdep splat in inet_csk_route_req()
lockdep_rcu_suspicious
inet_csk_route_req
tcp_v4_send_synack
tcp_rtx_synack
inet_rtx_syn_ack
tcp_fastopen_synack_time
tcp_retransmit_timer
tcp_write_timer_handler
tcp_write_timer
call_timer_fn
Thread running inet_csk_route_req() owns a reference on the request
socket, so we have the guarantee ireq->ireq_opt wont be changed or
freed.
lockdep can enforce this invariant for us.
Fixes: c92e8c02fe ("tcp/dccp: fix ireq->opt races")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When retransmission on TSQ handler was introduced in the commit
f9616c35a0 ("tcp: implement TSQ for retransmits"), the retransmitted
skbs' timestamps were updated on the actual transmission. In the later
commit 385e20706f ("tcp: use tp->tcp_mstamp in output path"), it stops
being done so. In the commit, the comment says "We try to refresh
tp->tcp_mstamp only when necessary", and at present tcp_tsq_handler and
tcp_v4_mtu_reduced applies to this. About the latter, it's okay since
it's rare enough.
About the former, even though possible retransmissions on the tasklet
comes just after the destructor run in NET_RX softirq handling, the time
between them could be nonnegligibly large to the extent that
tcp_rack_advance or rto rearming be affected if other (remaining) RX,
BLOCK and (preceding) TASKLET sofirq handlings are unexpectedly heavy.
So in the same way as tcp_write_timer_handler does, doing tcp_mstamp_refresh
ensures the accuracy of algorithms relying on it.
Fixes: 385e20706f ("tcp: use tp->tcp_mstamp in output path")
Signed-off-by: Koichiro Den <den@klaipeden.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use BUG_ON instead of if condition followed by BUG in do_setlink.
This issue was detected with the help of Coccinelle.
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Because SYSTEMPORT is a (semi) normal network device, the stack may attempt to
queue packets on it oustide of the DSA slave transmit path. When that happens,
the DSA layer has not had a chance to tag packets with the appropriate per-port
and per-queue information, and if that happens and we don't have a port 0 queue
0 available (e.g: on boards where this does not exist), we will hit a NULL
pointer de-reference in bcm_sysport_select_queue().
Guard against such cases by testing for the TX ring validity.
Fixes: 84ff33eeb23d ("net: systemport: Establish DSA network device queue mapping")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko says:
====================
mlxsw: Add support for non-equal-cost multi-path
Ido says:
In the device, nexthops are stored as adjacency entries in an array
called the KVD linear (KVDL). When a multi-path route is hit the
packet's headers are hashed and then converted to an index into KVDL
based on the adjacency group's size and base index.
Up until now the driver ignored the `weight` parameter for multi-path
routes and allocated only one adjacency entry for each nexthop with a
limit of 32 nexthops in a group. This set makes the driver take the
`weight` parameter into account when allocating adjacency entries.
First patch teaches dpipe to show the size of the adjacency group, so
that users will be able to determine the actual weight of each nexthop.
The second patch refactors the KVDL allocator, making it more receptive
towards the addition of another partition later in the set.
Patches 3-5 introduce small changes towards the actual change in the
sixth patch that populates the adjacency entries according to their
relative weight.
Last two patches finally add another partition to the KVDL, which allows
us to allocate more than 32 entries per-group and thus support more
nexthops and also provide higher accuracy with regards to the requested
weights.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The KVD linear is currently partitioned into two partitions. One for
single entries and another for groups of 32 entries.
Add another partition consisting of groups of 512 entries which will
allow us to more accurately represent the nexthop weights in non-equal
cost multi-path routing.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The memory region where adjacency entries (nexthops) are stored is
called the KVD linear and is configured during initialization with a
size of 64K.
Extend this area with 32K more entries, that will be partitioned into 64
groups of 0.5K entries, thereby allowing us to support weighted nexthops
with high accuracy.
Change the ratio between both types of hash entries, so as to prevent
reduction in the number of double hash entries, which are used for IPv6
neighbours and routes with a prefix length greater than 64.
Note that the user will be able to control all these sizes once the
devlink resource manager is introduced.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up until now the driver assumed all the nexthops have an equal weight
and wrote each to a single adjacency entry.
This patch takes the `weight` parameter into account and populates the
adjacency group according to the relative weight of each nexthop.
Specifically, the weights of all the nexthops that should be offloaded
are first normalized and then used to calculate the upper adjacency
index of each nexthop. This is done according to the hash-threshold
algorithm used by the kernel for IPv4 multi-path routing.
Adjacency groups are currently limited to 32 entries which limits the
weights that can be used, but follow-up patches will introduce groups of
512 entries.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>