The hardware (supposedly) has a 25ms timeout for clock stretching
and the driver uses 100ms which should be plenty.
The interrupt path however misses handling for errors while waiting for
the completion and the polling path uses an open-coded readx_poll_timeout.
Note that we drop reg_write(smbus, REG_SMSTA, status) while fixing those
issues here which will be done anyway whenever the next transaction starts
via pasemi_smb_clear.
Signed-off-by: Sven Peter <sven@svenpeter.dev>
Link: https://lore.kernel.org/r/20250427-pasemi-fixes-v3-2-af28568296c0@svenpeter.dev
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
The I2C bus can get stuck under some conditions (desync between
controller and device). The pasemi controllers include an unjam feature
that is enabled on reset, but was being disabled by the driver. Keep it
enabled by explicitly setting the UJM bit in the CTL register. This
should help recover the bus from certain conditions, which would
otherwise remain stuck forever.
Signed-off-by: Hector Martin <marcan@marcan.st>
Reviewed-by: Neal Gompa <neal@gompa.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Signed-off-by: Sven Peter <sven@svenpeter.dev>
Link: https://lore.kernel.org/r/20250427-pasemi-fixes-v3-1-af28568296c0@svenpeter.dev
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
Add functions to perform block read and write operations. This applies
for cases where the requested operation is for >8 bytes of data.
When not using the block mode transfer, the driver will attempt a series
of 8 byte i2c operations until it reaches the desired total. For
example, for a 40 byte request the driver will complete 5 separate
transactions. This results in large transactions taking a significant
amount of time to process.
Add block mode such that the driver can request larger transactions, up
to 1024 bytes per transfer.
Many aspects of the block mode transfer is common with the regular 8
byte operations. Use generic functions for parts of the message
construction and sending the message. The key difference for the block
mode is the usage of separate FIFO buffer to store data.
Write to this buffer in the case of a write (before command send).
Read from this buffer in the case of a read (after command send).
Data is written into this buffer by placing data into the MSB onwards.
This means the bottom 8 bits of the data will match the top 8 bits, and
so on and so forth.
Set specific bits in message for block mode, enable block mode transfers
from global i2c management registers, construct message, send message,
read or write from FIFO buffer as required.
The block-mode transactions result in a significant speed increase in
large i2c requests.
Signed-off-by: Aryan Srivastava <aryan.srivastava@alliedtelesis.co.nz>
Link: https://lore.kernel.org/r/20250324192946.3078712-2-aryan.srivastava@alliedtelesis.co.nz
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
The PCC driver now handles mapping and unmapping of shared memory
areas as part of pcc_mbox_{request,free}_channel(). Without these before,
this xgene-slimpro I2C driver did handling of those mappings like several
other PCC mailbox client drivers.
There were redundant operations, leading to unnecessary code. Maintaining
the consistency across these driver was harder due to scattered handling
of shmem.
Just use the mapped shmem and remove all redundant operations from this
driver.
Cc: Andi Shyti <andi.shyti@kernel.org>
Cc: linux-i2c@vger.kernel.org
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Link: https://lore.kernel.org/r/20250411112303.1149086-1-sudeep.holla@arm.com
Signed-off-by: Andi Shyti <andi.shyti@kernel.org>
irq_domain_add_linear() is going away as being obsolete now. Switch to
the preferred irq_domain_create_linear(). That differs in the first
parameter: It takes more generic struct fwnode_handle instead of struct
device_node. Therefore, of_fwnode_handle() is added around the
parameter.
Note some of the users can likely use dev->fwnode directly instead of
indirect of_fwnode_handle(dev->of_node). But dev->fwnode is not
guaranteed to be set for all, so this has to be investigated on case to
case basis (by people who can actually test with the HW).
[ tglx: Fix up subject prefix ]
Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20250319092951.37667-20-jirislaby@kernel.org
Prepare to resolve conflicts with an upstream series of fixes that conflict
with pending x86 changes:
6f5bf947ba Merge tag 'its-for-linus-20250509' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When the list_for_each_entry_reverse() exits without hitting a break
then the list cursor points to invalid memory. So this check for
if (c2a->fixed) is checking bogus memory. Fix it by using a "found"
variable to track if we found what we were looking for or not.
The list head (i2c_atr_chan.alias_pairs) is not a full entry, it's just
a struct list_head. When the for loop runs to completion, c2a doesn't
point to a struct i2c_atr_alias_pair, so you can't access c2a->fixed.
Fixes: c3f5524188 ("i2c: Support dynamic address translation")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Reviewed-by: Romain Gantois <romain.gantois@bootlin.com>
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
The i2c-atr module keeps a list of associations between I2C client aliases
and I2C addresses. This represents the address translation table which is
programmed into an ATR channel at any given time. This list is only updated
when a new client is bound to the channel.
However in some cases, an ATR channel can have more downstream clients than
available aliases. One example of this is an SFP module that is bound to an
FPC202 port. The FPC202 port can only access up to two logical I2C
addresses. However, the SFP module may expose up to three logical I2C
addresses: its EEPROM on 7-bit addresses 0x50 and 0x51, and a PHY
transceiver on address 0x56.
In cases like these, it is necessary to reconfigure the channel's
translation table on the fly, so that all three I2C addresses can be
accessed when needed.
As there are currently no known ATR's which do not support dynamic address
translation, this feature can be enabled by default without breaking
existing use cases.
Modify the i2c-atr module to provide on-the-fly address translation. This
is achieved by modifying an ATR channel's translation table whenever an I2C
transaction with unmapped clients is requested.
Add a mutex to protect alias_list. This prevents
i2c_atr_dynamic_attach/detach_addr from racing with the bus notifier
handler to modify alias_list.
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Romain Gantois <romain.gantois@bootlin.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Some I2C address translators (ATRs) assign each of their remote peripheral
aliases to a specific channel. To properly handle these devices, add
support for having separate alias pools for each ATR channel.
This is achieved by allowing callers of i2c_atr_add_adapter to pass an
optional alias list. If present, this list will be used to populate the
channel's alias pool. Otherwise, the common alias pool will be used.
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Romain Gantois <romain.gantois@bootlin.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
The "alias_list" field of struct i2c_atr_chan describes translation table
entries programmed in the ATR channel. This terminology will become more
confusing when per-channel alias pool support is introduced, as struct
i2c_atr_chan will gain a new field called "alias_pool", which will describe
aliases which are available to the ATR channel.
Rename the "alias_list" field to "alias_pairs" to clearly distinguish it
from the future "alias_pool" field.
No functional change is intended.
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Romain Gantois <romain.gantois@bootlin.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Each I2C address translator (ATR) has a pool of client aliases which can be
used as translation targets. Some ATRs have a single alias pool shared by
all downstream channels, while others have a separate alias pool for each
channel. Currently, this alias pool is represented by the "aliases",
"num_aliases", and "use_mask" fields of struct i2c_atr.
In preparation for adding per-channel alias pool support, move the
"aliases", "num_aliases", "use_mask" and associated lock to a new struct
called "struct alias_pool".
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Romain Gantois <romain.gantois@bootlin.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
The I2C Address Translator (ATR) module defines mappings from i2c_client
structs to aliases. However, only the physical address of each i2c_client
struct is actually relevant to the workings of the ATR module. Moreover,
some drivers require address translation functionality but do not allocate
i2c_client structs, accessing the adapter directly instead. The SFP
subsystem is an example of this.
Replace the "i2c_client" field of the i2c_atr_alias_pair struct with a u16
"addr" field. Rewrite helper functions and callbacks as needed.
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Tested-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Romain Gantois <romain.gantois@bootlin.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
