The original vendor's driver programmed the dma controller's
AHB HPROT values to enable bufferable, privileged mode. This
along with the "same priorty for both channels" fixes the
freezes according to @takimata, @And.short, that have been
reported on the forum by @ticerex.
Furtheremore, @takimata reported that the patch also improved
the performance of the HDDs considerably:
|<https://forum.lede-project.org/t/wd-mybook-live-duo-two-disks/16195/55>
|It seems your patch unleashed the full power of the SATA port.
|Where I was previously hitting a really hard limit at around
|82 MB/s for reading and 27 MB/s for writing, I am now getting this:
|
|root@OpenWrt:/mnt# time dd if=/dev/zero of=tempfile bs=1M count=1024
|1024+0 records in
|1024+0 records out
|real 0m 13.65s
|user 0m 0.01s
|sys 0m 11.89s
|
|root@OpenWrt:/mnt# time dd if=tempfile of=/dev/null bs=1M count=1024
|1024+0 records in
|1024+0 records out
|real 0m 8.41s
|user 0m 0.01s
|sys 0m 4.70s
|
|This means: 121 MB/s reading and 75 MB/s writing!
|
|[...]
|
|The drive is a WD Green WD10EARX taken from an older MBL Single.
|I repeated the test a few times with even larger files to rule out
|any caching, I'm still seeing the same great performance. OpenWrt is
|now completely on par with the original MBL firmware's performance.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Restarting service sysctl echos multiple errors like:
sysctl: -e: No such file or directory
After the first filename, all remaining arguments are treated
as files.
Signed-off-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
procd needs processes to stay in foreground to remain under its gaze and
control. Failure to do so means service stop commands fail to actually
stop the process (procd doesn't think it's running 'cos the process has
exited already as part of its forking routing)
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit 9d5a246930)
Drop the parallel-unsafe custom Build/dtb macro and use the .dtb artifacts
produced by the generic image build code.
Also remove unused .dtb references in the mt7623 subtarget.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
(backported from commit 8194f9ef4a)
First of all lengths should be compared after checking all blocks for
being good/bad. It's because requested length may differ from a final
one if there were some bad blocks.
Secondly it makes sense to also compare crc32 since we already have a
new one calculated.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 82498a7f7a)
Reading MTD data with (p)read doesn't return any error when accessing
bad block. As the result, with current code, CRC32 covers "data" stored
in bad blocks.
That behavior doesn't match CFE's one (bootloader simply skips bad
blocks) and may result in:
1) Invalid CRC32
2) CFE refusing to boot firmware with a following error:
Boot program checksum is invalid
Fix that problem by checking every block before reading its content.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 0f54489f75)
Bump to the latest cake recipe.
This backports tc class support to kernel 4.9 and other than conditional
kernel compilation pre-processor macros represents the cake that has
gone upstream into kernel 4.19. Loud cheer!
Fun may be had by changing cake tin classification for packets on
ingress. e.g.
tc filter add dev ifb4eth0 parent 800b: protocol ip u32 match \
ip dport 6981 0xffff action skbedit priority 800b:1
Where 800b: represents the filter handle for the ifb obtained by 'tc
qdisc' and the 1 from 800b:1 represents the cake tin number. So the
above example puts all incoming packets destined for port 6981 into the
BULK (lowest priority) tin.
f39ab9a Obey tin_order for tc filter classifiers
1e2473f Clean up after latest backport.
82531d0 Reorder includes to fix out of tree compilation
52cbc00 Code style cleanup
6cdb496 Fix argument order for NL_SET_ERR_MSG_ATTR()
cab17b6 Remove duplicate call to qdisc_watchdog_init()
71c7991 Merge branch 'backport-classful'
32aa7fb Fix compilation on Linux 4.9
9f8fe7a Fix compilation on Linux 4.14
ceab7a3 Rework filter classification
aad5436 Fixed version of class stats
be1c549 Add cake-specific class stats
483399d Use tin_order for class dumps
80dc129 Add class dumping
0c8e6c1 Fix dropping when using filters
c220493 Add the minimum class ops
5ed54d2 Start implementing tc filter/class support
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit c729c43b39)
The previous callback code was fragile, dependent on some UCI callback
bugs and side-effects now fixed in master commit 73d8a6ab.
Update scripts to use callbacks where appropriate and necessary, while
using normal UCI config parsing for all else. This results in smaller,
simpler, more robust code. Use callbacks in generate.sh to only process
'interface' defaults and the varying entries for 'reclassify', 'default'
and 'classify' sections. Also switch qos-stat to use non-callback UCI
handling.
The current changes work independently of 73d8a6ab (i.e. both before and
after), and are consistent with UCI config parsing documentation.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
There are several long-standing issues present in the UCI shell API as
documented in https://wiki.openwrt.org/doc/devel/config-scripting. They
relate both to high-level, user-defined callback functions used to
process UCI config files, and also to low-level functions used within
scripts generally.
The related problems have been encountered now and in the past, e.g.
https://forum.openwrt.org/viewtopic.php?id=54295, and include:
a) UCI parsing option() function and user-defined option_cb() callbacks
being erroneously called during processing of "list" config file entries;
b) normal usage of the low-level config_set() unexpectedy calling any
defined option_cb() if present; and
c) handling of the list_cb() not respecting the NO_CALLBACK variable.
Root causes include a function stack "inversion", where the low-level
config_set() function incorrectly calls the high-level option() function,
intended only for processing the "option" keyword of UCI config files.
This change addresses the inversion and other issues, making the option
handling code more consistent and smaller, and simplifying developers'
usage of UCI callbacks.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase]
Starting with 4.14, the "amcc,dwc-otg" needs to be used
in order to get the usb-otg to work.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
(cherry picked from commit 12b80f1cab)
Fix broken DHCPv6 servers which provide the server unicast option but
do not reply on DHCPv6 renew messages directed to the IPv6 address
contained in the server unicast option which results in broken IPv6
connectivity.
67ae6a7 odhcp6c: add option to ignore Server Unicast option
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
This commit adds the get_status_led method to diag.sh, which sets the
boot-led as status-led for scripts using this method to get a
status-led.
This method is used platform-independent in downstream project gluon to
set the LED used to indicate the config-mode.
Signed-off-by: David Bauer <mail@david-bauer.net>
(cherry picked from commit 4fc005197a)
On Debian, busybox does have a time applet, but it does not support the -f
flag. Catch this in prereq check to give users to proper error message.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit b123921a92)
An incorrect variable name was referenced in KERNEL_FILE_DEPENDS, leading
to the omission of the backport-* patch dirs in the generation of the
prepared stamp name.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit 36fa1bbf6f)
Backport hot off the press upstream netlink patch. Fixes stats display
from CAKE qdisc on MIPS allowing us to bump CAKE to latest version.
The gen_stats facility will add a header for the toplevel nlattr of type
TCA_STATS2 that contains all stats added by qdisc callbacks. A reference
to this header is stored in the gnet_dump struct, and when all the
per-qdisc callbacks have finished adding their stats, the length of the
containing header will be adjusted to the right value.
However, on architectures that need padding (i.e., that don't set
CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS), the padding nlattr is added
before the stats, which means that the stored pointer will point to the
padding, and so when the header is fixed up, the result is just a very
big padding nlattr. Because most qdiscs also supply the legacy TCA_STATS
struct, this problem has been mostly invisible, but we exposed it with
the netlink attribute-based statistics in CAKE.
Fix the issue by fixing up the stored pointer if it points to a padding
nlattr.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit 3698b34a00)
- remove misaligned custom buffer allocation in the NAND driver
- remove broken bounce buffer implementation for 16-byte align
Let the MTD core take care of both
Fixes messages like these:
[ 102.820541] Data buffer not 16 bytes aligned: 87daf08c
Signed-off-by: Felix Fietkau <nbd@nbd.name>
08719b1 mt76: use a per rx queue page fragment cache
4d2c565 mt76x2: reset HW before probe
f622975 mt76x2: fix CCK protection control frame rate
6780375 mt76x2: add frame protection support
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Using the NAPI or netdev frag cache along with other drivers can lead to
32 KiB pages being held for a long time, despite only being used for
very few page fragment.
This can happen if the ethernet driver grabs one or two fragments for rx
ring refill, while other drivers use (and free up) the remaining
fragments. The 32 KiB higher-order page can only be freed once all users
have freed their fragments, which only happens after the rings of all
drivers holding the fragments have wrapped around.
Depending on the traffic patterns, this can waste a lot of memory and
look a lot like a memory leak
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Allow enabling/commenting/disabling each feed individually by using a
tristate config symbol.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit 16035a7dd3)
FEEDS_ENABLED and FEEDS_DISABLED are derived from FEEDS_AVAILABLE, not
FEEDS_INSTALLED.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit 6dac434c00)
Setting CONFIG_FEED_... symbols combined two different effects: Disabling
a feed in the generated opkg distfeeds.conf, and omitting the feed from
PACKAGE_SUBDIRS.
It does not make sense to omit built feeds from PACKAGE_SUBDIRS, as it will
only lead to packages that can be enabled in .config (and that will
consequently be built) not to be found during rootfs creation, breaking
the build. All feeds that packages are emitted to should simply always be
added to PACKAGE_SUBDIRS instead; the CONFIG_FEED_... only configure the
generated distfeeds.conf like this.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit 9af22f1ac9)
The src-dummy method does not actually obtain any feed, but it can be used
to insert addtional entries into the opkg distfeeds.conf. This is useful to
make package feeds available to users without requiring the corresponding
source feeds to be available during build.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
(cherry picked from commit 6bdd5d8459)
The line that produces factory image was accidentally left by me while
testing before inital commit.
I came to the conclusion that flashing from OEM firmware does not work
(seems to share this behavior with other tplinks based on mt7628).
I have not done any further analysis, as I was unable to open the
case and attach a serial port (too much glue). Maybe i will try once
more.
So the way to do initial flashing (or un-bricking) is to use the
tftp-recover image. It is possible to revert to OEM firmware with tftp
recovery; in this case the first 512 bytes the image file need to be
cut off.
Signed-off-by: Peter Lundkvist <peter.lundkvist@gmail.com>
[add explaination provided via mail as commit message]
Signed-off-by: Mathias Kresin <dev@kresin.me>
Refreshed all patches
Reworked patches to match upstream:
335-v4.16-netfilter-nf_tables-add-single-table-list-for-all-fa.patch
Compile-tested on: cns3xxx, imx6, x86_64
Runtime-tested on: cns3xxx, imx6, x86_64
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Enables support for Dynack feature.
When a remote station is far away, we need to compensate for the distance
by allowing more time for an ACK to arrive back before issueing a retransmission.
Currently, it needs to be set fixed to indicate the maximum distance the remote
station will ever be.
While this mostly works for static antennae, it introduces 2 issues:
- If the actual distance is less, speed is reduced due to a lot of wates wait-time
- If the distance becomes greater, retries start to occur and comms can get lost.
Allowing to set it dynamically using dynack ensures the best possible tradeoff
between speed vs distance.
This feature is currently only supported in ath9k.
it is also disabled by default.
Enabling it can be done in 2 ways:
- issue cmd: iw phy0 set distance auto
- sending the NL80211_ATTR_WIPHY_DYN_ACK flag to mac80211 driver using netlink
Disabling it can be done by providing a valid fixed value.
To give an idea of a practical example:
In my usecase, we have mesh wifi device installed on ships/platforms.
Currently, the coverage class is set at 12000m fixed.
When a vessel moved closer (ex. 1500m), the measured link capacity was a lot
lower compared to setting the coverage class fixed to 1500m
Dynack completely solved this, nearly providing double the bandwidth at closer range
compared to the fixed setting of 12000m being used.
Also when a vessel sailed to a distance greater than the fixed setting,
communication was lost as the ACK's never arrived within the max allowed timeframe.
Actual distance: 6010m
iperf 60s run avg
Fixed 12150m: 31 Mbit/s
Dynack: 58 Mbit/s
Fixed 6300m: 51 Mbit/s
Dynack: 59 Mbit/s
Fixed 3000m: 13 Mbit/s (lots of retries)
Dynack: 58 Mbit/s
Actual distance: 1504m
iperf 60s run avg
Fixed 12150m: 31 Mbit/s
Dynack: 86 Mbit/s
Fixed 6300m: 55 Mbit/s
Dynack: 87 Mbit/s
Fixed 3000m: 67 Mbit/s
Dynack: 87 Mbit/s
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Partition name is picked by a parser_trx_data_part_name(). It has to
get correct partition offset (taking care of bad blocks) to work
properly.
This fixes UBI support for devices that have kernel flashed on partition
with a bad block.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit 7ddba08d87)
Most of the implementations behind cfg80211_get_station will not initialize
sinfo to zero before manipulating it. For example, the member "filled",
which indicates the filled in parts of this struct, is often only modified
by enabling certain bits in the bitfield while keeping the remaining bits
in their original state. A caller without a preinitialized sinfo.filled can
then no longer decide which parts of sinfo were filled in by
cfg80211_get_station (or actually the underlying implementations).
cfg80211_get_station must therefore take care that sinfo is initialized to
zero. Otherwise, the caller may tries to read information which was not
filled in and which must therefore also be considered uninitialized. In
batadv_v_elp_get_throughput's case, an invalid "random" expected throughput
may be stored for this neighbor and thus the B.A.T.M.A.N V algorithm may
switch to non-optimal neighbors for certain destinations.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
(cherry picked from commit 87493dac11)
Board Data File (BDF) is loaded upon driver boot-up procedure. The right
board data file is identified on QCA4019 using bus, bmi-chip-id and
bmi-board-id.
The problem, however, can occur when the (default) board data file cannot
fulfill the vendor requirements and it is necessary to use a different
board data file.
This problem was solved for SMBIOS by adding a special SMBIOS type 0xF8.
Something similar has to be provided for systems without SMBIOS but with
device trees. No solution was specified by QCA and therefore a new one has
to be found for ath10k.
The device tree requires addition strings to define the variant name
wifi@a000000 {
status = "okay";
qcom,ath10k-calibration-variant = "RT-AC58U";
};
wifi@a800000 {
status = "okay";
qcom,ath10k-calibration-variant = "RT-AC58U";
};
This would create the boarddata identifiers for the board-2.bin search
* bus=ahb,bmi-chip-id=0,bmi-board-id=16,variant=RT-AC58U
* bus=ahb,bmi-chip-id=0,bmi-board-id=17,variant=RT-AC58U
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
(cherry picked from commit 1c01e02575)
dfd9827 version: bump snapshot
88729f0 wg-quick: android: prevent outgoing handshake packets from being dropped
1bb9daf compat: more robust ktime backport
68441fb global: use fast boottime instead of normal boottime
d0bd6dc global: use ktime boottime instead of jiffies
18822b8 tools: fix misspelling of strchrnul in comment
0f8718b manpages: eliminate whitespace at the end of the line
590c410 global: fix a few typos
bb76804 simd: add missing header
7e88174 poly1305: give linker the correct constant data section size
fd8dfd3 main: test poly1305 before chacha20poly1305
c754c59 receive: don't toggle bh
Compile-tested-for: ath79 Archer C7 v2
Run-tested-on: ath79 Archer C7 v2
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit 3ce11588f6)
Fixes a potential infinite loop bug when in unlimited (ie not using
built in shaper) mode.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit 4bd4ece9ea)
This bumps to the latest & possibly greatest cake, sadly it's still
inedible but from an SQM point of view quite tasty :-)
Main tweaks since previous bump, improved ack_filter, some extra stats,
support for 64bit netlink parameters (higher rates/byte counters)
0520a6c Fix NAT option handling
8da93e1 Make sure we always call qdisc_watchdog_init() in cake_init()
f65daf6 Fix mismatched parenthesis
51d4ab3 Change flag handling to be safe even when mixing with non-eligible ACKs
f2ea091 ack_filter: protect DCTCP with stricter filtering of ECE marks
28b4560 ACK filter: Handle wrapping sequence numbers and DSACKs
73f62d9 Use the right PAD attribute for options
5969c14 Use 32 for tin backlog
e289f31 Move all the u64 netlink attributes together
36180a0 Check ACK seqno before parsing SACKs
91bbc01 Merge branch 'mine' into cobalt
58c55ec Rework SACK check to compare the ranges of two SACKs
9a5d593 ack_filter: Add proper handling of SACKs
eca95d4 ack_filter: short-circuit TCP flag check
d50a246 compat: backport some ktime functions
7b7ad11 compat: define tcpopt_fastopen for pre-4.1 kernels
ca54cdb Fix ktime compare
9d7dcc0 ack filter: Parse TCP options and only drop safe ones
b119882 Return EOPNOTSUPP on NAT option if conntrack is not available
842d7f0 Don't try to pad stats with tin_stats padding
bd46dc2 Use 64-bit divide helper
8e41bf0 Make sure we never drop SACKs when filtering ACKs
66e5d60 Avoid comparing ktime_t to scalar values
7fab017 Actually commit the ktime_t changes
fca6d13 Switch to ktime_t and get rid of cobalt.h
6f7e5af Can't use do_div with 64-bit divisors
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit abeae38dbb)
Bump iproute2/tc support of cake.
Add support for cake's change to u64 attribute passing for certain
attributes (rate & byte counts)
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
(cherry picked from commit a2165f936e)
Netgear R7800 switch LAN ports are numbered backwards in LuCI,
i.e. numbering is not corresponding to the actual physical port labels,
patch fixes that.
Signed-off-by: Aleksandr V. Piskunov <aleksandr.v.piskunov@gmail.com>
[merged with existing board using the same config]
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
To keep the status of a LED connected to the stp during boot, the get
callback is required. If the callback is missing and the LED default
state is set to keep in the devicetree, the gpio led driver errors out
during load.
Fixes: FS#1620
Signed-off-by: Mathias Kresin <dev@kresin.me>
Move SCAN_DEPS to scan.mk to eliminate redundancy with scripts/feeds
Add image/*.mk to SCAN_DEPS for targets to pick up newly added devices
Signed-off-by: Felix Fietkau <nbd@nbd.name>
In some cases, recent builds fail to boot from flash with at least some
MT7621 based devices. The error message is:
"LZMA ERROR 1 - must RESET board to recover"
Booting the same kernel via TFTP works for some reason.
Through testing I figured out that limiting the LZMA dictionary size
seems to prevent these errors
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Move to i2c pins pinmux node to the pinctrl node.
Fixes: a0685deec4 ("ramips: Add i2c support for mt7620n")
Signed-off-by: Andrey Jr. Melnikov <temnota.am@gmail.com>
[fix commit message]
Signed-off-by: Mathias Kresin <dev@kresin.me>
Felix Fietkau