Add "no-bus" configuration for regmap API
Subsystems (2)
cover Add "no-bus" configuration for regmap API
Second version of the initial patchset(can be found here:
https://lkml.org/lkml/2012/12/21/98)
This version contains following improvements:
- Provisions to switch between mutexes and spinlocks for locking in "no-bus" configuration
- "reg_read" and "reg_write" have "void *context" as a first argument in all the patches
- Minor improvements in documentation
Andrey Smirnov (3):
Add provisions to have user-defined read operation Add provisions to have user-defined write operation Add "no-bus" option for regmap API
drivers/base/regmap/internal.h | 5 ++ drivers/base/regmap/regmap.c | 156 +++++++++++++++++++++++++++++----------- include/linux/regmap.h | 18 ++++- 3 files changed, 135 insertions(+), 44 deletions(-)
Signature · 1 line
1.7.9.5
PATCH RESEND 1/3 Add provisions to have user-defined read operation
This commit is a preparatory commit to provide "no-bus" configuration option for regmap API. It adds necessary plumbing needed to have the ability to provide user define register read function.
--- drivers/base/regmap/internal.h | 2 ++ drivers/base/regmap/regmap.c | 35 ++++++++++++++++++++++++++--------- 2 files changed, 28 insertions(+), 9 deletions(-)
drivers/base/regmap/internal.h +2 -0
@@ -74,6 +74,8 @@ struct regmap {const struct regmap_access_table *volatile_table;const struct regmap_access_table *precious_table; +int (*reg_read)(void *context,unsigned int reg,unsigned int *val); +u8 read_flag_mask;u8 write_flag_mask;
drivers/base/regmap/regmap.c +26 -9
@@ -34,6 +34,9 @@ static int _regmap_update_bits(struct regmap *map, unsigned int reg, unsigned int mask, unsigned int val, bool *change); +static int _regmap_bus_read (void *context,unsigned int reg, +unsigned int *val); +bool regmap_reg_in_ranges (unsigned int reg,const struct regmap_range *ranges,unsigned int nranges)
@@ -430,6 +433,8 @@ struct regmap *regmap_init(struct device *dev, map->read_flag_mask = bus->read_flag_mask; } + map->reg_read = _regmap_bus_read; + reg_endian = config->reg_format_endian;if (reg_endian ==REGMAP_ENDIAN_DEFAULT ) reg_endian = bus->reg_format_endian_default;
@@ -1202,10 +1207,27 @@ static int _regmap_raw_read(struct regmap *map, unsigned int reg, void *val,return ret; } +static int _regmap_bus_read (void *context,unsigned int reg, +unsigned int *val) +{ +int ret; +struct regmap *map = context; + +if (!map->format.parse_val) +return -EINVAL ; + + ret =_regmap_raw_read (map, reg, map->work_buf, map->format.val_bytes); +if (ret ==0 ) + *val = map->format.parse_val (map->work_buf); + +return ret; +} +static int _regmap_read (struct regmap *map,unsigned int reg,unsigned int *val) {int ret; +BUG_ON (!map->reg_read);if (!map->cache_bypass) { ret =regcache_read (map, reg, val);
@@ -1213,26 +1235,21 @@ static int _regmap_read(struct regmap *map, unsigned int reg,return 0 ; } -if (!map->format.parse_val) -return -EINVAL ; -if (map->cache_only)return -EBUSY ; - ret =_regmap_raw_read (map, reg, map->work_buf, map->format.val_bytes); + ret = map->reg_read (map, reg, val);if (ret ==0 ) { - *val = map->format.parse_val (map->work_buf); -#ifdef LOG_DEVICE if (strcmp (dev_name (map->dev),LOG_DEVICE ) ==0 )dev_info (map->dev,"%x => %x\n" , reg, *val);#endif trace_regmap_reg_read (map->dev, reg, *val); - } -if (ret ==0 && !map->cache_bypass) -regcache_write (map, reg, *val); +if (!map->cache_bypass) +regcache_write (map, reg, *val); + }return ret; }
Signature · 1 line
1.7.9.5
Mark Brown
AppliedRe: [PATCH 1/3] [regmap] [RESEND] Add provisions to have user-defined read operation
Quoted 3 lines · Andrey Smirnov
This commit is a preparatory commit to provide "no-bus" configuration option for regmap API. It adds necessary plumbing needed to have the ability to provide user define register read function.
Applied, thanks.
PATCH RESEND 2/3 Add provisions to have user-defined write operation
In reply to Andrey SmirnovThis commit is a preparatory commit to provide "no-bus" configuration option for regmap API. It adds necessary plumbing needed to have the ability to provide user define register write function.
--- drivers/base/regmap/internal.h | 1 + drivers/base/regmap/regmap.c | 83 ++++++++++++++++++++++++++-------------- 2 files changed, 55 insertions(+), 29 deletions(-)
drivers/base/regmap/internal.h +1 -0
@@ -75,6 +75,7 @@ struct regmap {const struct regmap_access_table *precious_table;int (*reg_read)(void *context,unsigned int reg,unsigned int *val); +int (*reg_write)(void *context,unsigned int reg,unsigned int val);u8 read_flag_mask;u8 write_flag_mask;
drivers/base/regmap/regmap.c +54 -29
111 lines not on this page · raw message
Signature · 1 line
1.7.9.5
Mark Brown
AppliedRe: [PATCH 2/3] [regmap] [RESEND] Add provisions to have user-defined write operation
Quoted 3 lines · Andrey Smirnov
This commit is a preparatory commit to provide "no-bus" configuration option for regmap API. It adds necessary plumbing needed to have the ability to provide user define register write function.
Applied, thanks.
PATCH RESEND 3/3 Add "no-bus" option for regmap API
In reply to Andrey SmirnovThis commit adds provision for "no-bus" usage of the regmap API. In this configuration user can provide API with two callbacks 'reg_read' and 'reg_write' which are to be called when reads and writes to one of device's registers is performed. This is useful for devices that expose registers but whose register access sequence does not fit the 'bus' abstraction.
--- drivers/base/regmap/internal.h | 2 ++ drivers/base/regmap/regmap.c | 52 ++++++++++++++++++++++++++++++---------- include/linux/regmap.h | 18 +++++++++++++- 3 files changed, 59 insertions(+), 13 deletions(-)
drivers/base/regmap/internal.h +2 -0
@@ -77,6 +77,8 @@ struct regmap {int (*reg_read)(void *context,unsigned int reg,unsigned int *val);int (*reg_write)(void *context,unsigned int reg,unsigned int val); +bool cache_registers; +u8 read_flag_mask;u8 write_flag_mask;
drivers/base/regmap/regmap.c +40 -12
160 lines not on this page · raw message
include/linux/regmap.h +17 -1
@@ -127,7 +127,18 @@ typedef void (*regmap_unlock)(void *);* @lock_arg: this field is passed as the only argument of lock/unlock * functions (ignored in case regular lock/unlock functions * are not overridden). -* +* @reg_read: Optional callback that if filled will be used to perform +* all the reads from the registers. Should only be provided for +* devices whos read operation cannot be represented as a simple read +* operation on a bus such as SPI, I2C, etc. Most of the devices do +* not need this. +* @reg_write: Same as above for writing. +* @fast_io: Register IO is fast. Use a spinlock instead of a mutex +* to perform locking. This field is ignored if custom lock/unlock +* functions are used (see fields lock/unlock of struct regmap_config). +* This field is a duplicate of a similar file in +* 'struct regmap_bus' and serves exact same purpose. +* Use it only for "no-bus" cases. * @max_register: Optional, specifies the maximum valid register index. * @wr_table: Optional, points to a struct regmap_access_table specifying * valid ranges for write access.
@@ -177,6 +188,11 @@ struct regmap_config {regmap_unlock unlock;void *lock_arg; +int (*reg_read)(void *context,unsigned int reg,unsigned int *val); +int (*reg_write)(void *context,unsigned int reg,unsigned int val); + +bool fast_io; +unsigned int max_register;const struct regmap_access_table *wr_table;const struct regmap_access_table *rd_table;
Signature · 1 line
1.7.9.5
Re: [PATCH 3/3] [regmap] [RESEND] Add "no-bus" option for regmap API
drivers/base/regmap/internal.h struct regmap { · line 81
Quoted 2 lines · Andrey Smirnov
drivers/base/regmap/internal.h +2
-0+bool cache_registers; +
I'm afraid I don't quite understand this...
drivers/base/regmap/regmap.c int _regmap_write(struct regmap *map, unsigned int reg, · line 1064
Quoted 3 lines
drivers/base/regmap/regmap.c +1
-1-if (!map->cache_bypass && map->format.format_write) { +if (!map->cache_bypass && map->cache_registers) { ret =regcache_write (map, reg, val);
...I think it's mostly to service this check here, but looking at the code I can't quite think why the code is doing what it's doing - I suspect we should just remove the check for format_write() here. I think it was there to support potential bulk writes from the cache code (which we don't do yet) but we're not actually doing those and it's not clear that this should be doing that anyway.
drivers/base/regmap/regmap.c static int _regmap_read(struct regmap *map, unsigned int reg, · line 1278
Quoted 3 lines
drivers/base/regmap/regmap.c +2
-0int ret; +void *context = (map->bus) ? map : map->bus_context; +
Can you please make this a static inline regmap_map_get_context() or something? The same thing appears in quite a few places and the terery operator isn't great at the best of times.
Otherwise this looks good, thanks a lot for doing this.
Re: [PATCH 3/3] [regmap] [RESEND] Add "no-bus" option for regmap API
Quoted 19 lines · Mark Brown
drivers/base/regmap/internal.h struct regmap { · line 81
Quoted 2 lines · Andrey Smirnov
drivers/base/regmap/internal.h +2
-0+bool cache_registers; +I'm afraid I don't quite understand this...
drivers/base/regmap/regmap.c int _regmap_write(struct regmap *map, unsigned int reg, · line 1064
Quoted 3 lines
drivers/base/regmap/regmap.c +1
-1-if (!map->cache_bypass && map->format.format_write) { +if (!map->cache_bypass && map->cache_registers) { ret =regcache_write (map, reg, val);...I think it's mostly to service this check here, but looking at the code I can't quite think why the code is doing what it's doing - I suspect we should just remove the check for
format_write()here. I think it was there to support potential bulk writes from the cache code (which we don't do yet) but we're not actually doing those and it's not clear that this should be doing that anyway.
Quoted 1 line
From my understanding of the code it is done because the caching is
handled differently for cases when format_write() and format_reg(), format_val() are provided.
In case of 'format_write' the regcache_write() is called in _regmap_write() directly whereas when format_reg(), format_val() are there _regmap_write() calls _regmap_raw_write() which in turn calls regcache_write(). If I remove that variable and corresponding check then regcache_write() would be called twice in the case of format_reg(), format_val(), when _regmap_write() is called, would it not? I apologise if I miss something obvious and that is not a case(or issue).
Quoted 8 lines
drivers/base/regmap/regmap.c static int _regmap_read(struct regmap *map, unsigned int reg, · line 1278
Quoted 3 lines
int ret;drivers/base/regmap/regmap.c +2
-0+void *context = (map->bus) ? map : map->bus_context; +Can you please make this a static inline
regmap_map_get_context()or something? The same thing appears in quite a few places and the terery operator isn't great at the best of times.
Will do in the next version of this patch.
Completely unrelated off-topic question: the following code in regmap_bulk_write()
...
if (map->use_single_rw) {
for (i = 0; i < val_count; i++) {
ret = regmap_raw_write(map,
reg + (i * map->reg_stride),
val + (i * val_bytes),
val_bytes);
if (ret != 0)
return ret;
}
}....
Would it not deadlock, since both regmap_bulk_write() and regmap_raw_write() call map->lock(...)?
Once again, sorry for off-topic, I just didn't want to start a new thread because of what very well may by my poor understanding of the code(I'll submit a patch to change regmap_raw_write() to _regmap_raw_write() If I am correct).
Re: [PATCH 3/3] [regmap] [RESEND] Add "no-bus" option for regmap API
Quoted 11 lines · Andrey Smirnov
On Sun, Jan 13, 2013 at 3:18 PM, Mark Brown
Quoted 5 lines · Andrey Smirnov
Quoted 2 lines
+ bool
cache_registers;I'm afraid I don't quite understand this...
From my understanding of the code it is done because the caching is handled differently for cases when
format_write()andformat_reg(),format_val()are provided.
Quoted 8 lines
In case of '
format_write' theregcache_write()is called in_regmap_write()directly whereas whenformat_reg(),format_val()are there_regmap_write()calls_regmap_raw_write()which in turn callsregcache_write(). If I remove that variable and corresponding check thenregcache_write()would be called twice in the case offormat_reg(),format_val(), when_regmap_write()is called, would it not? I apologise if I miss something obvious and that is not a case(or issue).
OK, in this case the variable is confusingly named - it has no effect on if we're going to cache, it's about where we cache. What's really driving the decision here is a combination of having block I/O support (this was done this way to support cache for block writes) and having the ability to read (which is what limits us). Not sure I can think of a good name right now though...
Quoted 2 lines · Andrey Smirnov
Second version of the initial patchset(can be found here:
https://lkml.org/lkml/2012/12/21/98)
Please do use subject lines appropriate for the subsystem - in general, if your commit logs look different to others for the same subsystem there's probably an issue. Not doing this makes it easier for things to get missed. Here you want "regmap: "...