kernel/linux.git/drivers/memory/ti-aemif.c, branch v7.0.10

memory: ti-aemif: Export aemif_*_cs_timings()

2024-12-09T19:38:27+00:00

Export the aemif_set_cs_timing() and aemif_check_cs_timing() symbols so they can be used by other drivers Add a mutex to protect the CS configuration register from concurrent accesses between the AEMIF and its 'children'. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-7-bastien.curutchet@bootlin.com [krzysztof: wrap aemif_set_cs_timings() at 80-char] Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Create aemif_set_cs_timings()

2024-12-09T19:37:07+00:00

Create an aemif_set_cs_timings() function to isolate the setting of a chip select timing configuration and ease its exportation. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-6-bastien.curutchet@bootlin.com Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Create aemif_check_cs_timings()

2024-12-09T19:37:02+00:00

aemif_calc_rate() checks the validity of a new computed timing against a 'max' value given as input. This isn't convenient if we want to check the CS timing configuration somewhere else in the code. Wrap the verification of all the chip select's timing configuration into a single function to ease its exportation in upcoming patches. Remove the validity check from aemif_calc_rate(). Also remove the no longer used 'max' input and change the return type to u32. Remove the check of the aemif_calc_rate()'s return value during device-tree parsing as aemif_calc_rate() can't fail anymore. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-5-bastien.curutchet@bootlin.com Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Wrap CS timings into a struct

2024-12-09T19:37:01+00:00

CS timings are store in the struct aemif_cs_data along with other CS parameters. It isn't convenient for exposing CS timings to other drivers without also exposing the other parameters. Wrap the CS timings in a new struct aemif_cs_timings to simplify their export in upcoming patches. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-4-bastien.curutchet@bootlin.com Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Remove unnecessary local variables

2024-12-09T19:36:54+00:00

CS timings are copied to local variables that are then used as is, without any modifications. Remove these unneeded local variables and deal directly with the timings stored in the struct aemif_cs_data. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-3-bastien.curutchet@bootlin.com Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Store timings parameter in number of cycles - 1

2024-12-09T19:34:36+00:00

The CS configuration register expects timings to be expressed in 'number of cycles - 1' but they are stored in ns in the struct aemif_cs_data. So at init, the timings currently set are converted to ns by aemif_get_hw_params(), updated with values from the device-tree properties, and then converted back to 'number of cycles - 1' before being applied. Store the timings directly in 'number of cycles - 1' instead of nanoseconds. Perform the conversion from nanosecond during the device-tree parsing. Remove aemif_cycles_to_nsec() as it isn't used anymore. Signed-off-by: Bastien Curutchet Reviewed-by: Miquel Raynal Link: https://lore.kernel.org/r/20241204094319.1050826-2-bastien.curutchet@bootlin.com Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: Revert "memory: ti-aemif: don't needlessly iterate over child nodes"

2024-08-25T08:24:23+00:00

This reverts commit 23a641d5c2bce4c723fff9118a5d865ee6b9d05a. The first-level children of the aemif node are not the device nodes (ones containing the 'compatible' property) but the chip-select nodes which instead have their own children. of_platform_populate() will skip such nodes so we must indeed iterate over the direct children of the aemif node. The problem here is that we never call of_platform_depopulate() as it takes the root device as argument. We only have an unpopulated chip-select nodes so we will leak these devices if any of the calls to of_platform_populate() fails. I don't have a batter idea right now but my patch was not correct so we need to revert it. While at it: at least use the scoped variant of the OF node iterator. Down the line, we should find a better solution to fix this potential resource leak in error path. Signed-off-by: Bartosz Golaszewski Link: https://lore.kernel.org/r/20240824080235.56472-1-brgl@bgdev.pl Fixes: 23a641d5c2bc ("memory: ti-aemif: don't needlessly iterate over child nodes") Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: simplify with scoped for each OF child loop

2024-08-21T11:41:00+00:00

Use scoped for_each_available_child_of_node_scoped() when iterating over device nodes to make code a bit simpler. Reviewed-by: Jonathan Cameron Link: https://lore.kernel.org/r/20240816-cleanup-h-of-node-put-memory-v2-13-9eed0ee16b78@linaro.org Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: simplify with dev_err_probe()

2024-08-21T11:41:00+00:00

Use dev_err_probe() to avoid dmesg flood on actual defer. This makes the code also simpler. Reviewed-by: Jonathan Cameron Link: https://lore.kernel.org/r/20240816-cleanup-h-of-node-put-memory-v2-11-9eed0ee16b78@linaro.org Signed-off-by: Krzysztof Kozlowski

memory: ti-aemif: don't needlessly iterate over child nodes

2024-08-12T10:32:11+00:00

When populating devices from a specific node, we don't need to call of_platform_populate() for every child node manually - the routine will already do it. We can call it directly from the platform device's OF node. While at it: use the managed variant of of_platform_populate(). Signed-off-by: Bartosz Golaszewski Link: https://lore.kernel.org/r/20240809-ti-aemif-v1-3-27b1e5001390@linaro.org Signed-off-by: Krzysztof Kozlowski