Merge "Samsung DTS non-critical fixes for v4.7":
1. Fix s5p-mfc driver probe on Exynos542x Peach boards (need to provide MFC
memory banks). On these boards this was broken for long time but
apparently no one enabled this driver till now.
2. Fix creation of debugfs entries for one regulator on Exynos4210
Trats board.
3. Fix probing of max8997 MFD driver (and its children) because
of missing interrupt. Actually the current version of the driver probes
(just without interrupts) but after switching to regmap and regmap-irq,
the interrupt will be mandatory.
* tag 'samsung-fixes-non-critical-4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/krzk/linux:
ARM: dts: exynos: Add interrupt line to MAX8997 PMIC on exynos4210-trats
ARM: dts: exynos: Fix regulator name to avoid forbidden character on exynos4210-trats
ARM: dts: exynos: Add MFC memory banks for Peach boards
ARM: EXYNOS: Properly skip unitialized parent clock in power domain on
arm-ccn driver uses irq_set_affinity, which is not exported and
hence cannot be built as a module, eventhough we have all the
bits ready. This patch makes use of the exported helper
irq_set_affinity_hint() instead. Also, the __free_irq expects
the affinity_hint to be NULL when we free the irq. So set the
affinity_hint to NULL at clean up.
Now that we can build it as a module, update the Kconfig to
reflect the change.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Merge "Reset controller changes for v4.7" from Philipp Zabel:
- add missing stub for device_reset
- add support for OXNAS SoCs
* tag 'reset-for-4.7-2' of git://git.pengutronix.de/git/pza/linux:
reset: Add missing function stub for device_reset
dt-bindings: Add Oxford Semiconductor Reset Controller bindings
reset: Add Oxford Semiconductor Reset Controller driver
Merge "ARM: mediatek: kconfig updates for v4.7" from Matthias Brugger:
Add mt2701 support
* tag 'v4.6-next-kconfig' of https://github.com/mbgg/linux-mediatek:
ARM: mediatek: Add MT2701 config options for mediatek SoCs.
Merge "ARM: mediatek soc updates for v4.7" from Matthias Brugger:
- re-organize pmic wrapper code for easier and cleaner addiont of new SoCs and pmic wrappers
- add support for pmic wrapper mt6323
- add support for SoC mt2701
- enable gpt6 arch timer on mt7623
* tag 'v4.6-next-soc' of https://github.com/mbgg/linux-mediatek:
ARM: mediatek: enable gpt6 on boot up to make arch timer work on mt7623
soc: mediatek: PMIC wrap: add MT2701/7623 support
soc: mediatek: PMIC wrap: add mt6323 slave support
soc: mediatek: PMIC wrap: add a slave specific struct
soc: mediatek: PMIC wrap: remove pwrap_is_mt8135() and pwrap_is_mt8173()
soc: mediatek: PMIC wrap: move wdt_src into the pmic_wrapper_type struct
soc: mediatek: PMIC wrap: SPI_WRITE needs a different bitmask for MT2701/7623
soc: mediatek: PMIC wrap: WRAP_INT_EN needs a different bitmask for MT2701/7623
soc: mediatek: PMIC wrap: split SoC specific init into callback
soc: mediatek: PMIC wrap: add wrapper callbacks for init_reg_clock
soc: mediatek: PMIC wrap: don't duplicate the wrapper data
Merge "soc/tegra: Add generic PM domain support" from Thierry Reding:
Implements generic PM domain support on top of the existing Tegra power-
gate API. Drivers are thus allowed to move away from the Tegra-specific
API and towards using generic power domains directly.
* tag 'tegra-for-4.7-genpd' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
soc/tegra: pmc: Add generic PM domain support
dt-bindings: Add power domain info for NVIDIA PMC
Merge "ARM: tegra: Enable the XUSB controller" from Thierry Reding:
These changes add support for the XUSB controller on Tegra124. It is an
XHCI compatible controller that replaces the existing EHCI controllers.
Support is enabled on Venice2, Jetson TK1 and Nyan-based Chromebooks.
* tag 'tegra-for-4.7-xusb-no-defconfig' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
ARM: tegra: Enable XUSB on Nyan
ARM: tegra: Enable XUSB on Jetson TK1
ARM: tegra: Enable XUSB on Venice2
ARM: tegra: Add Tegra124 XUSB controller
ARM: tegra: Move Tegra124 to the new XUSB pad controller binding
This is a prerequisite for enabling the Tegra XUSB, all the
branches should be merged already at the time we get here.
* tegra/pci:
PCI: tegra: Support per-lane PHYs
dt-bindings: pci: tegra: Update for per-lane PHYs
phy: tegra: Add Tegra210 support
phy: Add Tegra XUSB pad controller support
dt-bindings: phy: tegra-xusb-padctl: Add Tegra210 support
dt-bindings: phy: Add NVIDIA Tegra XUSB pad controller binding
phy: core: Allow children node to be overridden
clk: tegra: Add interface to enable hardware control of SATA/XUSB PLLs
* tegra/usb:
usb: xhci: tegra: Add Tegra210 support
usb: xhci: Add NVIDIA Tegra XUSB controller driver
dt-bindings: usb: xhci-tegra: Add Tegra210 XUSB controller support
dt-bindings: usb: Add NVIDIA Tegra XUSB controller binding
phy: tegra: Add Tegra210 support
phy: Add Tegra XUSB pad controller support
dt-bindings: phy: tegra-xusb-padctl: Add Tegra210 support
dt-bindings: phy: Add NVIDIA Tegra XUSB pad controller binding
phy: core: Allow children node to be overridden
clk: tegra: Add interface to enable hardware control of SATA/XUSB PLLs
Merge "usb: host: xhci-tegra: Changes for v4.7-rc1" from Thierry Reding:
This set of patches introduces a driver for the XUSB controller found on
NVIDIA Tegra SoCs. When loaded with a firmware (available via the linux-
firmware repository), it provides an XHCI-compatible interface.
* tag 'tegra-for-4.7-xhci' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
usb: xhci: tegra: Add Tegra210 support
usb: xhci: Add NVIDIA Tegra XUSB controller driver
dt-bindings: usb: xhci-tegra: Add Tegra210 XUSB controller support
dt-bindings: usb: Add NVIDIA Tegra XUSB controller binding
Merge "PCI: tegra: Changes for v4.7-rc1" from Thierry Reding:
These patches update the Tegra PCIe host bridge controller device tree
bindings and driver to cope with per-lane PHYs on Tegra124 and later.
* tag 'tegra-for-4.7-pci' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
PCI: tegra: Support per-lane PHYs
dt-bindings: pci: tegra: Update for per-lane PHYs
Merge "phy: tegra: Changes for v4.7-rc1" from Thierry Reding:
This set of patches adds support for the Tegra XUSB pad controller. The
controller provides a set of pads (lanes) that are used for I/O by other
IP blocks within Tegra SoCs (PCIe, SATA and XUSB).
* tag 'tegra-for-4.7-phy' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux:
phy: tegra: Add Tegra210 support
phy: Add Tegra XUSB pad controller support
dt-bindings: phy: tegra-xusb-padctl: Add Tegra210 support
dt-bindings: phy: Add NVIDIA Tegra XUSB pad controller binding
phy: core: Allow children node to be overridden
clk: tegra: Add interface to enable hardware control of SATA/XUSB PLLs
Merge "Rockchip driver updates for v4.7 - part2" from Heiko Stübner:
Ability to save and restore the power-domain quality of service
settings that get lost and reset on power-domain power cycles.
* tag 'v4.7-rockchip-drivers-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip:
soc: rockchip: power-domain: support qos save and restore
dt-bindings: modify document of Rockchip power domains
Merge "aspeed arch tag for 4.7" from Joel Stanley <joel@jms.id.au>:
This is the initial pull request for the Aspeed BMC SoCs.
I put 4.7 in the subject for this and the subsequent requests but if it's a bit
late in the cycle then I understand.
Following review I got rid of the board file by pushing the functionality out
to a watchdog and clock driver, so it's just the Kconfig bits. I've also added
myself to maintainers for the Aspeed arch and drivers as I intend on looking
after them.
* tag 'aspeed-for-4.7-arch' of https://github.com/shenki/linux:
arm: Add Aspeed machine
Add myself as co-maintainer, update mailing list entry and add a couple
more directories.
Signed-off-by: Kevin Hilman <khilman@baylibre.com>
Acked-by: Carlo Caione <carlo@caione.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Merge "DaVinci documentation update for v4.7" from Sekhar Nori:
MAINTAINERS file update to fix some stale entries.
* tag 'davinci-for-v4.7/doc' of git://git.kernel.org/pub/scm/linux/kernel/git/nsekhar/linux-davinci:
MAINTAINERS: fix stale TI DaVinci entries
Merge "The i.MX SoC updates for 4.7" from Shawn Guo":
- Allow TWD to be used on UP kernel, as the PREEMPT-RT and cyclictest
shows that TWD has a slightly better performance than i.MX timer.
* tag 'imx-soc-4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/shawnguo/linux:
ARM: imx: always use TWD on IMX6Q
Previously when everything happened in the set_rate callbacks itself we
needed the old_rate value for the possible rate rollback, so that made
it easy to also use it in the debug output.
Now with the param-handling being done in separate functions, reading and
recalculating the current pll rate only to use it in a debug message that
won't get displayed in regular cases anyway is quite a waste.
Therefore drop that value from the debug output. In the worst case that
previous rate will have been displayed on the rate change before.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
With the previous commit, the clock drivers now know at init time if the
GRF regmap is available. That means if it isn't available then, it also
won't become available later and we can therefore switch PLLs, that need
the GRF for the lock-status, to read-only mode - similar behaviour as the
aborting of rate changes we did before.
This saves some conditionals on every rate change and we can also drop
the rockchip_clk_get_grf function completely.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
In the distant past syscons were initialized pretty late and weren't
available at the time the clock init ran. As the GRF is mainly needed
for PLL lock-status checking, we had this lazy init that tried to grab
the syscon on PLL rate changes and denied these changes if it was not
available.
These days syscons are available very early and recent addition to
rockchip clocks, like the PLL clk_init actually also rely on them
being available at that time, so there is no need to keep that lazy
init around, as it will also result in some more simplifications in
other parts of the clock-code.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Merge "mvebu soc for 4.7" from Gregory CLEMENT:
- Clock framework cleanup with the "Remove CLK_IS_ROOT" series
* tag 'mvebu-soc-4.7-1' of git://git.infradead.org/linux-mvebu:
ARM: dove: Remove CLK_IS_ROOT
ARM: orion5x: Remove CLK_IS_ROOT
ARM: mv78xx0: Remove CLK_IS_ROOT
Merge "DaVinci SoC updates for v4.7 (part 2)" from Sekhar Nori:
These patches provide ability to add non-PSC
clocks to DaVinci clock framwork and are
required to support USB PHY clock setting from
USB PHY driver.
* tag 'davinci-for-v4.7/soc-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/nsekhar/linux-davinci:
ARM: davinci: da850: use clk->set_parent for async3
ARM: davinci: Move clock init after ioremap.
Merge "NXP LPC32xx mach updates for v4.7" from Vladimir Zapolskiy:
This includes a few nonfunctional clean-ups for NXP LPC32xx:
* removed leftover from restart code migration to a watchdog driver
* removed dead code leftovers from migration to CCF driver
* fix double const qualifier
* tag 'lpc32xx-soc-4.7' of git://github.com/vzapolskiy/linux-lpc32xx:
ARM: lpc32xx: remove duplicate const on lpc32xx_auxdata_lookup
ARM: lpc32xx: remove leftovers of legacy clock source and provider drivers
ARM: lpc32xx: remove reboot header file
The change adds a list of files for maintenance under NXP LPC32xx
section, the listed files are NXP LPC32xx SoC series mach files,
DTS files of NXP LPC32xx SoC powered boards and NXP LPC32xx SoC
peripheral drivers, most of the peripheral driver file names match
'lpc32xx' pattern.
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Roland Stigge added initial support of NXP LPC32xx SoC series and
first boards powered by it, but for a while since v3.18-rc1 the
SoC support is unmaintained and became stale.
Vladimir Zapolskiy and Sylvain Lemieux expressed interest in
continuation of NXP LPC32xx maintenance, reflect this in MAINTAINERS
record file for better communication with Linux kernel community.
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
A record of NXP LPC32xx SoC support is lost between LMxx hwmon drivers
and lockdep, rename and move it to a place where all other ARM
SoC and machines settle.
Note, NXP LPC32xx maintenance is actually about SoC series itself, SoC
peripherals and a number of machines powered by LPC32xx SoC, so while
we are here correct the title name to emphasize that the maintenance
concerns SoC support in general.
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Now that we know exactly which page sizes our caller wants to use in the
given domain, we can restrict higher-order allocation attempts to just
those sizes, if any, and avoid wasting any time or effort on other sizes
which offer no benefit. In the same vein, this also lets us accommodate
a minimum order greater than 0 for special cases.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Tested-by: Yong Wu <yong.wu@mediatek.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Many IOMMUs support multiple page table formats, meaning that any given
domain may only support a subset of the hardware page sizes presented in
iommu_ops->pgsize_bitmap. There are also certain use-cases where the
creator of a domain may want to control which page sizes are used, for
example to force the use of hugepage mappings to reduce pagetable walk
depth.
To this end, add a per-domain pgsize_bitmap to represent the subset of
page sizes actually in use, to make it possible for domains with
different requirements to coexist.
Signed-off-by: Will Deacon <will.deacon@arm.com>
[rm: hijacked and rebased original patch with new commit message]
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
As a set of driver-provided callbacks and static data, there is no
compelling reason for struct iommu_ops to be mutable in core code, so
enforce const-ness throughout.
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The priv field from iommu_ops is a hangover from the of_dma_configure
series and isn't actually used. Remove it before it has chance to
spread.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When audio and music units have some quirks in their sequence of packet,
it's really hard for non-owners to identify the quirks. Although developers
need dumps for sequence of packets, it's difficult for users who have no
knowledges and no equipments for this purpose.
This commit adds tracepoints for this situation. When users encounter
the issue, they can dump a part of packet data via Linux tracing framework
as long as using drivers in ALSA firewire stack.
Additionally, tracepoints for outgoing packets will be our help to check
and debug packet processing of ALSA firewire stack.
This commit newly adds 'snd_firewire_lib' subsystem with 'in_packet' and
'out_packet' events. In the events, some attributes of packets and the
index of packet managed by this module are recorded per packet.
This is an usage:
$ trace-cmd record -e snd_firewire_lib:out_packet \
-e snd_firewire_lib:in_packet
/sys/kernel/tracing/events/snd_firewire_lib/out_packet/filter
/sys/kernel/tracing/events/snd_firewire_lib/in_packet/filter
Hit Ctrl^C to stop recording
^C
$ trace-cmd report trace.dat
...
23647.033934: in_packet: 01 4073 ffc0 ffc1 00 000f0040 9001b2d1 122 44
23647.033936: in_packet: 01 4074 ffc0 ffc1 00 000f0048 9001c83b 122 45
23647.033937: in_packet: 01 4075 ffc0 ffc1 00 000f0050 9001ffff 002 46
23647.033938: in_packet: 01 4076 ffc0 ffc1 00 000f0050 9001e1a6 122 47
23647.035426: out_packet: 01 4123 ffc1 ffc0 01 010f00d0 9001fb40 122 17
23647.035428: out_packet: 01 4124 ffc1 ffc0 01 010f00d8 9001ffff 002 18
23647.035429: out_packet: 01 4125 ffc1 ffc0 01 010f00d8 900114aa 122 19
23647.035430: out_packet: 01 4126 ffc1 ffc0 01 010f00e0 90012a15 122 20
(Here, some common fields are omitted so that a line to be within 80
characters.)
...
One line represent one packet. The legend for the last nine fields is:
- The second of cycle scheduled for the packet
- The count of cycle scheduled for the packet
- The ID of node as source (hex)
- Some devices transfer packets with invalid source node ID in their CIP
header.
- The ID of node as destination (hex)
- The value is not in CIP header of packets.
- The value of isochronous channel
- The first quadlet of CIP header (hex)
- The second quadlet of CIP header (hex)
- The number of included quadlets
- The index of packet in a buffer maintained by this module
This is an example to parse these lines from text file by Python3 script:
\#!/usr/bin/env python3
import sys
def parse_ts(second, cycle, syt):
offset = syt & 0xfff
syt >>= 12
if cycle & 0x0f > syt:
cycle += 0x10
cycle &= 0x1ff0
cycle |= syt
second += cycle // 8000
cycle %= 8000
# In CYCLE_TIMER of 1394 OHCI, second is represented in 8 bit.
second %= 128
return (second, cycle, offset)
def calc_ts(second, cycle, offset):
ts = offset
ts += cycle * 3072
# In DMA descriptor of 1394 OHCI, second is represented in 3 bit.
ts += (second % 8) * 8000 * 3072
return ts
def subtract_ts(minuend, subtrahend):
# In DMA descriptor of 1394 OHCI, second is represented in 3 bit.
if minuend < subtrahend:
minuend += 8 * 8000 * 3072
return minuend - subtrahend
if len(sys.argv) != 2:
print('At least, one argument is required for packet dump.')
sys.exit()
filename = sys.argv[1]
data = []
prev = 0
with open(filename, 'r') as f:
for line in f:
pos = line.find('packet:')
if pos < 0:
continue
pos += len('packet:')
line = line[pos:].strip()
fields = line.split(' ')
datum = []
datum.append(fields[8])
syt = int(fields[6][4:], 16)
# Empty packet in IEC 61883-1, or NODATA in IEC 61883-6
if syt == 0xffff:
data_blocks = 0
else:
payload_size = int(fields[7], 10)
data_block_size = int(fields[5][2:4], 16)
data_blocks = (payload_size - 2) / data_block_size
datum.append(data_blocks)
second = int(fields[0], 10)
cycle = int(fields[1], 10)
start = (second << 25) | (cycle << 12)
datum.append('0x{0:08x}'.format(start))
start = calc_ts(second, cycle, 0)
datum.append("0x" + fields[5])
datum.append("0x" + fields[6])
if syt == 0xffff:
second = 0
cycle = 0
tick = 0
else:
second, cycle, tick = parse_ts(second, cycle, syt)
ts = calc_ts(second, cycle, tick)
datum.append(start)
datum.append(ts)
if ts == 0:
datum.append(0)
datum.append(0)
else:
# Usual case, or a case over 8 seconds.
if ts > start or start > 7 * 8000 * 3072:
datum.append(subtract_ts(ts, start))
if ts > prev or start > 7 * 8000 * 3072:
gap = subtract_ts(ts, prev)
datum.append(gap)
else:
datum.append('backward')
else:
datum.append('invalid')
prev = ts
data.append(datum)
sys.exit()
The data variable includes array with these elements:
- The index of the packet
- The number of data blocks in the packet
- The value of cycle count (hex)
- The value of CIP header 1 (hex)
- The value of CIP header 2 (hex)
- The value of cycle count (tick)
- The value of calculated presentation timestamp (tick)
- The offset between the cycle count and presentation timestamp
- The elapsed ticks from the previous presentation timestamp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In callback function of isochronous context, modules can queue packets to
indicated isochronous cycles. Although the cycle to queue a packet is
deterministic by calculation, this module doesn't implement the calculation
because it's useless for processing.
In future, the cycle count is going to be printed with the other parameters
for debugging. This commit is the preparation. The cycle count is computed
by cycle unit, and correctly arranged to corresponding packets. The
calculated count is used in later commit.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
In callback function of isochronous context, u32 variable is passed for
cycle count. The value of this variable comes from DMA descriptors of 1394
Open Host Controller Interface (1394 OHCI). In the specification, DMA
descriptors transport lower 3 bits for second field and full cycle field in
16 bits field, therefore 16 bits of the u32 variable are available. The
value for second is modulo 8, and the value for cycle is modulo 8,000.
Currently, ALSA firewire-lib module don't use the value of the second
field, because the value is useless to calculate presentation timestamp in
IEC 61883-6. However, the value may be useful for debugging. In later
commit, it will be printed with the other parameters for debugging.
This commit makes this module to handle the whole cycle count including
second. The value is calculated by cycle unit. The existed code is already
written with ignoring the value of second, thus this commit causes no
issues.
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
There's no guarantee that an IP address in a different network namespace
actually represents the same endpoint.
Also, if we allow unprivileged nfs mounts some day then this might allow
an unprivileged user in another network namespace to misdirect somebody
else's nfs mounts.
If sharing between containers is really what's wanted then that could
still be arranged explicitly, for example with bind mounts.
Reported-by: "Eric W. Biederman" <ebiederm@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
At Connectathon 2016, we found that recent upstream Linux clients
would occasionally send a LOCK operation with a zero stateid. This
appeared to happen in close proximity to another thread returning
a delegation before unlinking the same file while it remained open.
Earlier, the client received a write delegation on this file and
returned the open stateid. Now, as it is getting ready to unlink the
file, it returns the write delegation. But there is still an open
file descriptor on that file, so the client must OPEN the file
again before it returns the delegation.
Since commit 24311f8841 ('NFSv4: Recovery of recalled read
delegations is broken'), nfs_open_delegation_recall() clears the
NFS_DELEGATED_STATE flag _before_ it sends the OPEN. This allows a
racing LOCK on the same inode to be put on the wire before the OPEN
operation has returned a valid open stateid.
To eliminate this race, serialize delegation return with the
acquisition of a file lock on the same file. Adopt the same approach
as is used in the unlock path.
This patch also eliminates a similar race seen when sending a LOCK
operation at the same time as returning a delegation on the same file.
Fixes: 24311f8841 ('NFSv4: Recovery of recalled read ... ')
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
[Anna: Add sentence about LOCK / delegation race]
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
A mirror can be shared between multiple layouts, even with different
iomodes. That makes stats gathering simpler, but it causes a problem
when we get different creds in READ vs. RW layouts.
The current code drops the newer credentials onto the floor when this
occurs. That's problematic when you fetch a READ layout first, and then
a RW. If the READ layout doesn't have the correct creds to do a write,
then writes will fail.
We could just overwrite the READ credentials with the RW ones, but that
would break the ability for the server to fence the layout for reads if
things go awry. We need to be able to revert to the earlier READ creds
if the RW layout is returned afterward.
The simplest fix is to just keep two sets of creds per mirror. One for
READ layouts and one for RW, and then use the appropriate set depending
on the iomode of the layout segment.
Also fix up some RCU nits that sparse found.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
We're just as likely to have allocation problems here as we would if we
delay looking up the credential like we currently do. Fix the code to
get a rpc_cred reference early, as soon as the mirror is set up.
This allows us to eliminate the mirror early if there is a problem
getting an rpc credential. This also allows us to drop the uid/gid
from the layout_mirror struct as well.
In the event that we find an existing mirror where this one would go, we
swap in the new creds unconditionally, and drop the reference to the old
one.
Note that the old ff_layout_update_mirror_cred function wouldn't set
this pointer unless the DS version was 3, but we don't know what the DS
version is at this point. I'm a little unclear on why it did that as you
still need creds to talk to v4 servers as well. I have the code set
it regardless of the DS version here.
Also note the change to using generic creds instead of calling
lookup_cred directly. With that change, we also need to populate the
group_info pointer in the acred as some functions expect that to never
be NULL. Instead of allocating one every time however, we can allocate
one when the module is loaded and share it since the group_info is
refcounted.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
In later patches, we're going to want to allow the creds to be updated
when we get a new layout with updated creds. Have this function take
a reference to the cred that is later put once the call has been
dispatched.
Also, prepare for this change by ensuring we follow RCU rules when
getting a reference to the cred as well.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
All the callers already call that function before calling into here,
so it ends up being a no-op anyway.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Sometimes we might have a RCU managed credential pointer and don't want
to use locking to handle it. Add a function that will take a reference
to the cred iff the refcount is not already zero. Callers can dereference
the pointer under the rcu_read_lock and use that function to take a
reference only if the cred is not on its way to destruction.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Add function rpc_lookup_generic_cred, which allows lookups of a generic
credential that's not current_cred().
[jlayton: add gfp_t parm]
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
We need to be able to call the generic_cred creator from different
contexts. Add a gfp_t parm to the crcreate operation and to
rpcauth_lookup_credcache. For now, we just push the gfp_t parms up
one level to the *_lookup_cred functions.
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
An xdr_buf with head[0].iov_len = 0 and page_len = 0 will cause
xdr_init_decode() to incorrectly setup the xdr_stream. Specifically,
xdr->end is never initialized.
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Commit ea2cf22 created nfs_commit_info and saved &inode->i_lock inside
this NFS specific structure. This obscures the usage of i_lock.
Instead, save struct inode * so later it's clear the spinlock taken is
i_lock.
Should be no functional change.
Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Like other resend paths, mark the (old) hdr as NFS_IOHDR_REDO. This
ensures the hdr completion function will not count the (old) hdr
as good bytes.
Also, vector the error back through the hdr->task.tk_status like other
retry calls.
This fixes a bug with the FlexFiles layout where libaio was reporting more
bytes read than requested.
Signed-off-by: Weston Andros Adamson <dros@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Use devm_mfd_add_devices() for MFD devices registration and get
rid of .remove callback to remove MFD child-devices. This is done
by managed device framework.
Signed-off-by: Laxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
