Petr Machata [Tue, 6 Apr 2021 16:15:39 +0000 (18:15 +0200)]
Mellanox: Add new mlxsw_spectrum firmware xx.2008.2438
Add latest verified version of Mellanox Spectrum-family switch firmware,
for Spectrum (13.2008.2438), Spectrum-2 (29.2008.2438) and Spectrum-3
(30.2008.2438).
This update fixes the following issues:
- An issue where strictly-prioritized traffic gets a 9usec latency penalty
on a congested port.
- Attempts to configure DSCP prioritization (QPDM register) on CPU
port (local port 0) would result in FW crash.
- An issue in sampling traffic to the CPU due to incorrect trap
prioritization.
- An issue in bulk-reading flow counters.
- Several issues related to FW upgrade and in-service FW upgrade.
- Incomplete clearing of activity bits in ECMP containers with a very high
number of individual ECMP entries (100K).
- A crash during router deinit.
Signed-off-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Jeremy Linton [Thu, 8 Apr 2021 03:25:54 +0000 (22:25 -0500)]
brcm: Link CM4's WiFi firmware with DMI machine name.
The kernel when booted in ACPI mode with SMBIOS/DMI
data uses the DMI data to set the machine
manufacture/model.
The WiFi driver uses the machine string to select the
correct firmware image. Because of this, an additional
link to select the correct firmware is needed.
Cc: Hans de Goede <hdegoede@redhat.com> Cc: Peter Robinson <pbrobinson@gmail.com> Cc: Josh Boyer <jwboyer@kernel.org> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Hans de Goede [Thu, 25 Mar 2021 22:15:47 +0000 (23:15 +0100)]
rtl_bt: Add rtl8723bs_config-OBDA0623.bin symlink
Some x86/ACPI devices such as the Acer Aspire Switch 10E (SW3-016),
use an alternative ACPI hardware-id (HID) for the BT part of the
RTL8723BS chip.
Instead of using the standard OBDA8723 HID this device uses a HID
of OBDA0623. Despite using a different HID it does work fine with
the default config settings from the rtl8723bs_config-OBDA8723.bin
file. Add a symlink to that file, to fix BT not working on this
device due to a missing config file.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Hans de Goede [Thu, 25 Mar 2021 22:15:46 +0000 (23:15 +0100)]
brcm: Add nvram for the Chuwi Hi8 (CWI509) tablet
Add a NVRAM file for the brcmfmac43430a0 wifi/bit chip found on the
Chuw Hi8 (CWI509) tablet. This is based on copy of the existing
brcm/brcmfmac43430a0-sdio.jumper-ezpad-mini3.txt NVRAM file with some
the calibration values adjusted for the Chuwi Hi8 tablet.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Hans de Goede [Thu, 25 Mar 2021 22:15:45 +0000 (23:15 +0100)]
brcm: Add nvram for the Predia Basic tablet
Add a NVRAM file for the brcmfmac43340 wifi/bit chip found on the
predia-basic tablet. This is based on copy of the existing
brcm/brcmfmac43340-sdio.meegopad-t08.txt NVRAM file with some the
calibration values adjusted for the Predia Basic tablet.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Petr Machata [Wed, 24 Feb 2021 15:39:07 +0000 (16:39 +0100)]
Mellanox: Add new mlxsw_spectrum firmware xx.2008.2406
Add latest verified version of Mellanox Spectrum-family switch firmware,
for Spectrum (13.2008.2406), Spectrum-2 (29.2008.2406) and Spectrum-3
(30.2008.2406).
This release contains the following fixes and improvements:
- MIRROR_SAMPLER_ACTION.mirror_probability_rate inverted.
- When adjacency is replaced-if-inactive (RATR.opcode=3), bad parameter was
reported when replacing an active entry.
- TC pool occupancy watermark not cleared after port splits.
- Per-TC counter of ECN-marked packets supported.
- When duplicating VXLAN packets for head-end replication, number of
lookups was limited to 254, while Spectrum>=2 support 4K.
- Deadlock on port activation when 3 1x ports are active in a single port
cluster on Spectrum-3.
- A storm of congestion threshold events generated on stably-congested
traffic when a 200 Gbps port is flapped on Spectrum-3.
- Incorrect early scheduling parameters for 50G 2-way split on Spectrum-2.
- Timeouts when accessing certain CRspace addresses.
- Enable sampling trapped packets.
Signed-off-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Jérôme Pouiller [Mon, 8 Feb 2021 12:53:48 +0000 (13:53 +0100)]
silabs: add new firmware for WF200
Add version 3.12.1 (released the 2021-02-18) of firmware for Silabs
WF200. This firmware is necessary for the wfx driver currently hosted in
the staging/ directory of the Linux kernel.
Here the last modifications (only modifications tagged [LMAC] have an
impact on Linux):
New Features/Improvements:
[1725] - [UMAC] Wi-Fi power mode API with 'listen_interval = 0' is
taken into account
Bug fixes:
[1710] - [UMAC] Increase WPA3 connection authentication timeout
[1712] - [LMAC] Avoid beacon collision in concurrent mode
[1726] - [LMAC] Fix 'incoherent status and result_flags' warning
[1730] - [UMAC] Fix incorrect packet abort leading to assert
Petr Machata [Wed, 27 Jan 2021 14:41:24 +0000 (15:41 +0100)]
Mellanox: Add new mlxsw_spectrum firmware xx.2008.2304
Add latest verified version of Mellanox Spectrum-family switch firmware,
for Spectrum (13.2008.2304), Spectrum-2 (29.2008.2304) and Spectrum-3
(30.2008.2304).
This release fixes the following issues (among others):
- Rx pause packet reaction latency on Spectrum-3
- Certain high-numbered flow counters would not work on Spectrum-2.
- FW freeze on Spectrum-1 when under heavy PTP traffic load.
- Mirroring traffic to the CPU on Spectrum-2.
- Shared headroom would not be exhausted before drops started occurring.
- Per-priority Rx discard counters were not always cleared.
- 400-Gbps ports with PFC enabled kept sending pause frames without buffer
pressure.
- Several issues in flow control and traffic admittance, and FW freeze
after an on-the-fly shared buffer reconfiguration.
- A number of issues in forwarding line speed traffic to partially split
8x ports on Spectrum-3.
- A number of issues related to in-service FW upgrade (ISSU / ISFU).
- Port shaper influenced ingress rate when PFC was enabled.
- FW freeze after a port was set admin down when unrelated ports were
congested.
And includes the following new feature:
- Improvements to counter read performance
Signed-off-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Jeremy Linton [Mon, 11 Jan 2021 23:18:50 +0000 (17:18 -0600)]
brcm: Link RPi4's WiFi firmware with DMI machine name.
The kernel when booted in ACPI mode with SMBIOS/DMI
data uses that data to set the machine manufacture/model.
The RPi's WiFi driver uses the machine string to
select the correct firmware image. Because of
this, an additional link to select the correct
firmware is needed.
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Reviewed-by: Peter Robinson <pbrobinson@gmail.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Peter Robinson [Fri, 8 Jan 2021 10:00:29 +0000 (10:00 +0000)]
brcm: Update Raspberry Pi 3B+/4B NVRAM for downstream changes
The Raspberry Pi Foundation NVRAM files have been updated for a new
BT/WiFi coexistence parameter for firmwares newer than 177, now
we have new firmwares (7.45.221) from Cypress for the BCM43455
update the NVRAM for these parameters.
Update the RPi4B NVRAM boardflags3 parameter to match too, which
also happens to be what we already have for the RPi3B+.
Jeremy Linton [Fri, 8 Jan 2021 20:44:20 +0000 (14:44 -0600)]
cypress: Fix link direction
It appears the link direction was incorrect resulting
in dangling symlinks. This also causes the brcm devices
to be unable to find their firmware
Fixes: 060ad8b3d505e ("cypress: Link the new cypress firmware to the old brcm files") Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> Cc: Peter Robinson <pbrobinson@gmail.com> Cc: Hans de Goede <hdegoede@redhat.com> Cc: Chi-Hsien Lin <chi-hsien.lin@cypress.com> Cc: Josh Boyer <jwboyer@kernel.org> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Peter Robinson [Tue, 5 Jan 2021 16:04:42 +0000 (16:04 +0000)]
cypress: Link the new cypress firmware to the old brcm files
By linking the new firmware to the old file it allows exisiting
kernels to make use of the new firmware and make use of the
newer firmware with all the associated fixes.
Signed-off-by: Peter Robinson <pbrobinson@gmail.com> Cc: Hans de Goede <hdegoede@redhat.com> Cc: Chi-Hsien Lin <chi-hsien.lin@cypress.com> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Remove these old ones as they're likely vulnerable to at least the
KrØØk vulnerability (CVE-2019-15126).
Signed-off-by: Peter Robinson <pbrobinson@gmail.com> Cc: Hans de Goede <hdegoede@redhat.com> Cc: Chi-Hsien Lin <chi-hsien.lin@cypress.com> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Stefan Seyfried [Wed, 16 Dec 2020 13:24:40 +0000 (14:24 +0100)]
make AP6212 in bananpi m2 plus/zero work
BananaPi M2 zero and M2 plus boards contain an AP6212 module which
works well with the default AP6212 config. Create symlinks so that
they work out of the box.
Signed-off-by: Stefan Seyfried <seife+dev@b1-systems.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Petr Machata [Mon, 23 Nov 2020 16:22:12 +0000 (17:22 +0100)]
Mellanox: Add new mlxsw_spectrum firmware xx.2008.2018
Add latest verified version of Mellanox Spectrum-family switch firmware,
for Spectrum (13.2008.2018), Spectrum-2 (29.2008.2018) and Spectrum-3
(30.2008.2018).
This release fixes the following issues (among others):
- Prioritization of trapped control traffic on Spectrum-2 and Spectrum-3.
- Several edge cases where the FW could get stuck on Spectrum-2 and
Spectrum-3.
- FW flash issues on Spectrum-3
- Apparent resource exhaustion on Spectrum-3 due to wrong fencing.
- When trapping dropped packets from several TCs, they would only get
reported under one TC.
- Incorrect rejection of RIF counters with indices over 16 bits.
- An issue where port split might fail after port saw heavy traffic.
- Certain large policer CIR caused effective zero CIR.
- A race that would cause drops due to lack of buffer space.
And includes the following new features:
- Support for shared port headroom
- A new trap for L2 IPv6 DHCP traffic
- On Spectrum-2 and Spectrum-3, support ACL actions that perform ALU
operations between packet fields, immediate values and general-purpose
registers
- Early support for 8-way port split on Spectrum-3
Signed-off-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Josh Boyer <jwboyer@kernel.org>
Reto Schneider [Sun, 8 Nov 2020 23:42:54 +0000 (00:42 +0100)]
rtlwifi: v88.2 firmware files for RTL8192CU
The vendor driver rtl8188C_8192C_usb_linux_v4.0.1_6911.20130308 includes
new firmware files. These were extracted from data statements in that
driver to form these files.
Before this update, with version 80 of the firmware, the USB interface
of the RTL8192CU WLAN controller often locked itself up:
usb 1-2: device descriptor read/64, error -110
usb 1-2: device not accepting address 4, error -110
usb 1-2: device not accepting address 5, error -110
usb usb1-port2: unable to enumerate USB device
usb 2-2: device descriptor read/64, error -110
usb 2-2: device descriptor read/64, error -110
On ARMv5 based GARDENA smart gateways running Linux 4.19.78, this can
be reliably reproduced by rebooting (warm) the gateway multiple times
(max. 50 attempts needed).
Unlike users having this issues on a USB Wi-Fi dongle, resetting of the
chip by replugging is not an option on this gateway due to the lack of
any power cut functionality. Therefore, a (cold) reboot of the whole
gateway is needed.
Updating the firmware of the RTL8192CU WLAN controller from version
v80.0 to v88.2 (as per output of rtl8xxxu) resolves this issue.
The problem did no show up anymore for 1000 restarts.
Please note:
- Only rtl8192cufw_TMSC.bin tested (mainly on rtl8xxxu)
- rtl8192cu seems to work as well as before, but I can not rule out
that this new firmware version brings unwanted changes.
The Realtek drivers containing v88.2 of the firmware
(v4.0.1_6911.20130308 to v4.0.9_25039.20171107) have some changes
compared to the version v3.4.2_3727.20120404, for which I do not know
if those should be reflected in rtl8192cu.
Unrelated of the initially described USB problem, another issue still
remains after updating the firmware: Using the rtl8192cu driver,
scanning for available SSIDs yields no more results after a few hundred
scans (iw wlan0 scan). rtl8xxxu does not suffer from this problem.
Signed-off-by: Reto Schneider <code@reto-schneider.ch> Acked-by: Larry Finger <Larry.Finger@lwfinger.net> Tested-by: Chris Chiu <chiu@endlessos.org> Signed-off-by: Josh Boyer <jwboyer@kernel.org>