Some Ubiquiti U-boot versions, in particular the "U-Boot 1.1.4.2-s956
(Jun 10 2015 - 10:54:50)" found with AirOS 5.6, do not correctly flush the
caches for the whole kernel address range after decompressing the kernel
image, leading to hard to debug boot failures, depending on kernel version
and configuration.
As a workaround, prepend the relocate-kernels loader, which will invalidate
the caches after moving the kernel to the correct load address.
Reported-by: Andreas Ziegler <dev@andreas-ziegler.de>
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Package postinst will pass even names of builtin modules to insert_modules,
leading to postinst failing with error 255. This has been fixed in master
in r5279, but for lede-17.01 this minimal change is preferable.
Fixes FS#645, FS#893.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
6609e98 libfstools: add "const" to char pointer arguments in mount_move()
95c07d5 libfstools: fix foreachdir() to pass dir with a trailing slash
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
mbedtls changed in version 2.7.0 and 2.7.2 the soversion of the
libmbedcrypto.so library, use the old version again to be able to use
the new library with binaries compiled against the old mbedtls library.
Some binaries got rebuild to for the 2.7.0 release and are now using
libmbedcrypto.so.1, the older ones are still using libmbedcrypto.so.0.
Go back to libmbedcrypto.so.0 and make the system rebuild the binaries
which were rebuild for 2.7.0 again.
This should make the libmbedcrypto.so library be compatible with the old
version shipped with 17.01.
Fixes: 3ca1438ae0 ("mbedtls: update to version 2.7.2")
Fixes: f609913b5c ("mbedtls: update to version 2.7.0")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This is important fix for flash parsing in some corner cases. In case
of TRX subpartition with rootfs being aligned to the flash block size it
was incorrectly registered twice. Detecting & registering it as a
standalone partition was resulting in an incorrect "firmware" partition
size and possibly broken sysupgrade.
It wasn't noticed before because "rootfs" alignment depends on a kernel
size. It can happen though - depending on the configuration and the
kernel size.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
(cherry picked from commit f5195e72c0)
Previously, tplink_pharos_check_image() would accept any image with ELF
magic and only non-printable data in the support-list, as in this case the
while-read loop would not run at all. Add the new support-list offset and
ensure an image is only accepted when the model string is actually found.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Switching from kernel 4.4.120 to 4.4.124 introduced a regression in
the genirq code. It was caused by a commit 9d0273bb1c4b6 ("genirq: Use
irqd_get_trigger_type to compare the trigger type for shared IRQs").
On bcm53xx it breaks serial console and results in a flood of:
[ 22.078829] genirq: Flags mismatch irq 18. 00000080 (serial) vs. 00000080 (gpio)
[ 22.086432] genirq: Flags mismatch irq 18. 00000080 (serial) vs. 00000080 (gpio)
[ 22.601150] genirq: Flags mismatch irq 18. 00000080 (serial) vs. 00000080 (gpio)
[ 22.608845] genirq: Flags mismatch irq 18. 00000080 (serial) vs. 00000080 (gpio)
Later in the upstream "linux-4.4.y" branch that commit was reverted and
it was followed by a 4.4.126 release. Until we switch from 4.4.124 to
4.4.126 (or newer), let's backport that reverting commit.
Fixes: bed0ee7cbf ("Kernel: bump 4.4 to 4.4.124 for 17.01")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
This fixes the following security problems:
* CVE-2018-0488: Risk of remote code execution when truncated HMAC is enabled
* CVE-2018-0487: Risk of remote code execution when verifying RSASSA-PSS signatures
This release is also ABI incompatible with the previous one, but it is
API compatible.
Some functions used by a lot of other software was renamed and the old
function names are provided as a static inline now, but they are only
active when deprecated functions are allowed, deactivate the removal of
deprecated functions for now.
Also increase the PKG_RELEASE version to force a rebuild and update of
packages depending on mbedtls to handle the changed ABI.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The e2fsprogs host build fails on a glibc 2.27 host with
make[6]: Entering directory 'build_dir/host/e2fsprogs-1.43.7/debugfs'
CC create_inode.o
./../misc/create_inode.c:399:18: error: conflicting types for 'copy_file_range'
static errcode_t copy_file_range(ext2_filsys fs, int fd, ext2_file_t e2_file,
^~~~~~~~~~~~~~~
In file included from ./../misc/create_inode.c:19:0:
/usr/include/unistd.h:1110:9: note: previous declaration of 'copy_file_range' was here
ssize_t copy_file_range (int __infd, __off64_t *__pinoff,
^~~~~~~~~~~~~~~
Backport upstream commit "misc: rename copy_file_range to
copy_file_chunk" 01551bdba16ab16512a01affe02ade32c41ede8a in order to
fix this.
Signed-off-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
At least on some devices, LEDs don't work anymore since kernel 4.4.120.
Revert the broken change.
See also: https://www.spinics.net/lists/stable/msg223656.html
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
The default fragment low/high thresholds are 3 and 4 MB. On devices with
only 32MB RAM, these settings may lead to OOM when many fragments that
cannot be reassembled are received. Decrease fragment low/high thresholds
to 384 and 512 kB on devices with less than 64 MB RAM.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Copying ./src/* would fail when src exists, but is empty or only contains
hidden files.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
/usr/lib/opkg/status must not be removed completely, otherwise the
packages' conffile lists will be missing. Replace it with a reduced version
only containing the conffile entries.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
When a user removes a preinstalled opkg package, the package's prerm script
(and in particular our default_prerm) should run.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Many packages use the opkg conffiles field to list configuration files that
are to be retained on upgrades. Make this work on systems without opkg.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Compiling the Intel microcode package results in a
microcode.bin and a microcode-64.bin. As we can
decide based on the subtarget which should be used,
we'll only split the required .bin file with
iucode-tool.
x64 will get the intel-microcode-64.bin
All other variants will get intel-microcode.bin
The microcodes will be updated from preinit via a common
script - that's the earliest place where we can do it.
Signed-off-by: Zoltan HERPAI <wigyori@uid0.hu>
Use the Debian repository for sourcing the ucode files.
Current (20171205) includes support for fam17h CPUs already.
The microcodes will be updated from preinit via a common
script - that's the earliest place where we can do it.
Signed-off-by: Zoltan HERPAI <wigyori@uid0.hu>
Add tool to "compile" Intel microcode files. The tool will be
compiled for host (to split the microcode.dat) and for target
(to forcibly reload the microcode if required).
Instead of using the large microcode.bin/microcode-64.bin, the
splitted ucode files (separate for CPU families) will be
installed.
Signed-off-by: Zoltan HERPAI <wigyori@uid0.hu>
There was a mismatch between indicating factory reset and code actually
starting it. After 5 seconds status LED started blinking rapidly letting
user know it's ready to release reset button. In practice button had to
stay pressed for another second in order to relly start the process.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Fixes issue FS#1355.
LPAE extensions are enabled, but the A13 does not support them.
The result is the boot process stopping at "Starting kernel ..."
Fixes: 468735c3a2 ("target: sunxi: enable kvm support")
Signed-off-by: Matteo Scordino <matteo.scordino@gmail.com>
Looking for a wrong LED file name was stopping this code from find any
LED. This affects devices with only a red/amber power LED.
Fixes: 3aaee1ba02 ("bcm53xx: failsafe support")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
This significantly improves mountd stability & reliability by:
1) Sending hotplug.d event when appropriate
2) Properly unmounting
3) Handling corner cases when unmounting fails
4) Improving log messages
5f2c419 mount: drop duplicated includes
aaf2743 mount: call hotplug-call with ACTION=remove before trying to unmount
97da4ed mount: try lazy unmount if normal one fails
1b62489 mount: create not working symlink when unmounting fails
e77dc6d mount: reorder deleting code in the mount_enum_drives()
76766ae mount: rename tmp variables in the mount_add_list()
04b897f mount: drop duplicated rmdir() call from the mount_enum_drives()
a27ea3f mount: drop duplicated unlink() call from the mount_dev_del()
bf7cc33 mount: fix/improve unmounting log messages
36f9197 mount: fix removing mount point if it's expired
ed4270f mount: struct mount: replace "mounted" and "ignore" fileds with a "status"
1af9ca2 mount: change mount_dev_del() argument to struct mount *
7c8fea8 mount: rename /proc/mount parser to mount_update_mount_list()
7aadd1c mount: improve handling mounts table size
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Allow building perf on uncommon targets again.
Depending on the kernel version, not all of these archs will actually use
libunwind in perf. Still, it seems simpler and less error-prone to use the
same list that is defined in the libunwind package.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Works around two incompatiblities between glibc and (POSIX-compliant) musl:
- missing register definitions from asm/ptrace.h
- non-POSIX-compliant ucontext_t on PPC32 with glibc
Compile tested on mpc85xx.
Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
The current implementation only checked if uqmi itself executed
correctly which is also the case when the returned value is actually
an error.
Rework this, checking that CID is a numeric value, which can only
be true if uqmi itself also executed correctly.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
This bumps the 4.4. kernel in LEDE 17.01 to 4.4.116.
More Meltdown & Spectre mitigation.
* Refresh patches.
* Refresh x86/config for RETPOLINE.
* Deleted 8049-PCI-layerscape-Add-fsl-ls2085a-pcie-compatible-ID.patch (accepted upstream)
* Deleted 8050-PCI-layerscape-Fix-MSG-TLP-drop-setting.patch (accepted upstream)
* 650-pppoe_header_pad.patch does not apply anymore (code was replaced).
Bumps from 4.4.113 to 4.4.115 were handled by Kevin Darbyshire-Bryant.
Compile-tested on: ar71xx, ramips/mt7621, x86/64
Run-tested on: ar71xx, ramips/mt7621, x86/64
Signed-off-by: Stijn Segers <foss@volatilesystems.org>
uqmi contains a command for directly querying the modem if there
is a valid data connection, so let's use it.
This avoids the cases were all previous tests are succesful, but the
actual data link is not up for some reasons, leading to states were we
thought the link was up when it actually wasn't ..
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Originally, the implementation only checked if uqmi command
execution succeeded properly without actually checking it's returned data.
This lead to a pass, even when the returned data was indicating an error.
Rework the verification to actually check the returned data,
which can only be correct if the uqmi command itself also executed correctly.
On command execution success, value "pdh_" is a pure numeric value.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Debugging shows that using the general method properly cleans on each
run, while the method specifying the client-ID shows "No effect"
even while in connected state.
Fixes several connectivity issues seen on specific modems.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
If a device only supports the 2nd verification method (uim),
the first method will fail as expected reporting an error:
"Command not supported"
Silence both separate methods and only report an error regarding
pin verification if both fail.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Some newer LTE modems, like the MC7455 or EC25-E do not support
"802.3" mode, and will stay in "raw-ip" regardless of the mode being
set.
In this case, the driver must be informed that it should handle all
packets in raw mode. [1]
This commit fixes connectivity issues for these devices.
Before:
[ Node 5 ] udhcpc -i wwan0
udhcpc: started, v1.27.2
udhcpc: sending discover
udhcpc: sending discover
udhcpc: sending discover
After:
[ Node 5 ] udhcpc -i wwan0
udhcpc: started, v1.27.2
udhcpc: sending discover
udhcpc: sending select for 100.66.245.226
udhcpc: lease of 100.66.245.226 obtained, lease time 7200
udhcpc: ifconfig wwan0 100.66.245.226 netmask 255.255.255.252 broadcast
+
udhcpc: setting default routers: 100.66.245.225
[1] https://lists.freedesktop.org/archives/libqmi-
devel/2017-January/002064.html
Tested on cns3xxx using a Sierra Wireless MC7455 LTE-A
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
[bumped PKG_RELEASE]
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
If you unplug a QMI device, the /dev/cdc-wdmX device
disappears but uqmi will continue to poll it endlessly.
Then, when you plug it back, you have 2 uqmi processes,
and that's bad, because 2 processes talking QMI to the
same device [and the same time] doesn't seem to work well.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Backport support for raw-ip mode including all known fixes afterwards.
Newer LTE modems only tend to support this mode, which was only
introduced in kernel 4.5.
Also backport support for the Quectel EC2x LTE modem series which is
a very popular device.
No custom changes were needed in order to apply these patches.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Backport commits fixing the detection of GPT partition names during
preinit and sysupgrade, closing a shell-injection vulnerability.
da52dd0c83 ("base-files: quote values when evaluating uevent")
267873ac9b ("base-files: don't evaluate block-device uevent")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Kevin Darbyshire-Bryant