BMC/Intel-BMC/linux.git, branch dev-5.10.49-intel

i2c: aspeed: fix interrupt handling on some abnormal cases

2021-07-26T18:53:22+00:00

These two exceptional cases were rarely observed in a slave device reset or in a peer master reset conditions. 1. aspeed-i2c-bus 1e78a480.i2c-bus: irq handled != irq. expected 0x00000030, but was 0x00000020 when master goes from ASPEED_I2C_MASTER_STOP to ASPEED_I2C_MASTER_INACTIVE state due to an ASPEED_I2CD_INTR_ABNORMAL event. In this case, ASPEED_I2CD_INTR_NORMAL_STOP should be simply ignored without printing out the error message. 2. aspeed-i2c-bus 1e78a480.i2c-bus: irq handled != irq. expected 0x00000010, but was 0x00000000 when master already went to ASPEED_I2C_MASTER_INACTIVE. H/W sends one additional ASPEED_I2CD_INTR_NORMAL_STOP event but it's a garbage because driver is already in inactive state so filter it off from the error message printing out. Signed-off-by: Jae Hyun Yoo Change-Id: I94794a9612a7e8db445591e490bad6adae0faacf

fixup! i2c: Add mux hold/unhold msg types

2021-07-26T18:22:43+00:00

Fixup an unexpected bus holding case caused by a message sending failure on packet which has a holding message. In case of an error on sending a message, a bus holding should be immediately released by itself to prevent a race condition because user space doesn't send an unhold message when a packet which has hold message fails. This case was obversed in PLDM stress test which can make intensive AIC resets so driver was in a severe condition that has lots of arbitration losses and abnormal stops. Signed-off-by: Jae Hyun Yoo Change-Id: Ibd169c791d0c24f41932ea676d4e187ddbd3e140

mfd: intel-peci-client: Extend MFD cell id with additional parameter

2021-07-14T17:10:33+00:00

If we're only using CPU number as MFD cell id, we don't have a way to avoid a collision in case we have multiple adapters able to communicate with the same CPUs. [ 794.916872] sysfs: cannot create duplicate filename '/bus/platform/devices/peci-cputemp.0' [ 795.452507] intel_peci_client 1-30: Failed to register child devices: -17 [ 795.460238] intel_peci_client: probe of 1-30 failed with error -17 To fix this problem, let's use both adapter number and cpu number to calculate MFD cell id. Change-Id: If0559a0b629081e6201b4ce93cb8bc649dd2a258 Signed-off-by: Iwona Winiarska

ASD Disable JTAG Master controller output when driver is released

2021-07-14T17:10:33+00:00

To avoid conflicts with other JTAG hardware probes such as XDP, ASD must disable JTAG Master controller output when ASD connection is not in use. This commit fixes the output disable for AST26xx series which needs ASPEED_JTAG_CTRL and ASPEED_JTAG_GBLCTRL to be cleared for all SW, HW1 and HW2 modes. Test: 1. Run jtag_test 2. Run ASD Sanity(using XDP) - Failed here before 3. Run jtag_test 4. Run ASD Sanity(using ASD) 5. Run ASD Sanity(using XDP) - Failed here before 6. ASD Sanity(using ASD) Signed-off-by: Ernesto Corona Change-Id: I4f403f4b8412ab96cee195412cc412360845b6d0

hwmon: peci: clear time window configuration

2021-07-14T17:10:33+00:00

This patch makes time window setting to be cleared when disabling power capping just like others settings. Change-Id: I3a5fc9c1eb6a5787d131bf69cc18a5dae94a99f0 Signed-off-by: Zbigniew Lukwinski

hwmon: peci: timestamps for sensors correction

2021-07-14T17:10:33+00:00

This patch makes timestaps catching to happen just after PECI transaction. It makes reading more accurate since PECI request can hang while waiting in PECI driver queue. Change-Id: I1a356caa59148a5cebcfcdf89211038bfadb40d3 Signed-off-by: Zbigniew Lukwinski

Add chip unique id reading interface

2021-07-14T17:10:33+00:00

This commit adds an interface for reading chip unique id value. Optionally, the id can be encrypted using a dts-supplied hash data. Signed-off-by: Jae Hyun Yoo Signed-off-by: Vernon Mauery Signed-off-by: Arun P. Mohanan Change-Id: Ifd98500ea87b3d40e1738e583e077a74851fae35

soc: aspeed: mctp: Reject packets with invalid payload size

2021-07-14T17:10:33+00:00

When we program TX command register, it requires packet data size in dwords. Since the aspeed-mctp driver doesn't implement MCTP protocol, it copies MCTP packet "as it" form userspace and uses size from write() syscall to program TX command. If the size from write() doesn't match the payload length in packet PCIe VDM header, it causes MCTP HW to stop working and we are not able to reset it without platform power cycle. To avoid HW issues, let's verify if the data size from write() matches the payload length in PCIe VDM header. Signed-off-by: Iwona Winiarska Change-Id: I6da14babadcc65cb2ba4a2b685495d1baa92c169

Flash layout change for SPDM's AFM

2021-07-14T17:10:33+00:00

This change adds the SPI region needed for AFM(Attestation Firmware Manifest) which is used by SPDM Requester in PFR RoT. The change realigns the remaining flash region and appends the 256KB AFM region to staging region and reserves 768KB region for PSU and HSBP firmware. Tested: BMC boot in non-PFR and PFR environment. Change-Id: I322eebbdbb5dcffbfc45b8b50650ba4aaa5af186 Signed-off-by: Vikram Bodireddy

misc: aspeed: add a raw value setting interface for uart routing

2021-07-14T17:10:33+00:00

This commit adds a raw value setting interface for uart routing to provide an atomic way of switching the route. Signed-off-by: Jae Hyun Yoo Change-Id: Id8b80f44f9a503d9f25a692211fc5ec78500ce68