wrappers for fpga_manager_ops
Subsystems (3)
[email protected]
reviewerv4 cover wrappers for fpga_manager_ops
From: Tom Rix <[email protected]>
As followup from
https://lore.kernel.org/linux-fpga/[email protected]/
Boards should not be required to have noop functions. So improve or create fpga-mgr wrappers for the fpga_manager_ops. Remove the noop functions.
Refactor fpga-mgr to use the wrappers.
Changes from
v1:
commit subject,log
v2:
rebase to next-20210623
v3:
remove mops check
add write_sg wrapper
drop 'fpga-mgr: collect wappers and change to inline'
Tom Rix (7):
fpga-mgr: wrap the write_init() op
fpga-mgr: make write_complete() op optional
fpga-mgr: wrap the write() op
fpga-mgr: wrap the status() op
fpga-mgr: wrap the state() op
fpga-mgr: wrap the fpga_remove() op
fpga-mgr: wrap the write_sg() op
drivers/fpga/dfl-fme-mgr.c | 6 -- drivers/fpga/fpga-mgr.c | 111 +++++++++++++++++++++++------------ drivers/fpga/stratix10-soc.c | 6 -- drivers/fpga/ts73xx-fpga.c | 6 -- drivers/fpga/zynqmp-fpga.c | 7 --- 5 files changed, 75 insertions(+), 61 deletions(-)
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 1/7 fpga-mgr: wrap the write_init() op
From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a write_init() op if there is nothing for it do. So add a wrapper and move the op checking. Default to success.
--- drivers/fpga/fpga-mgr.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-)
drivers/fpga/fpga-mgr.c +12 -3
@@ -25,6 +25,15 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline int fpga_mgr_write_init (struct fpga_manager *mgr, +struct fpga_image_info *info, +const char *buf,size_t count) +{ +if (mgr->mops->write_init) +return mgr->mops->write_init (mgr, info, buf, count); +return 0 ; +} +/** * fpga_image_info_alloc - Allocate an FPGA image info struct * @dev: owning device
@@ -83,9 +92,9 @@ static int fpga_mgr_write_init_buf(struct fpga_manager *mgr, mgr->state =FPGA_MGR_STATE_WRITE_INIT ;if (!mgr->mops->initial_header_size) - ret = mgr->mops->write_init (mgr, info,NULL ,0 ); + ret =fpga_mgr_write_init (mgr, info,NULL ,0 );else - ret = mgr->mops->write_init ( + ret =fpga_mgr_write_init ( mgr, info, buf,min (mgr->mops->initial_header_size, count));if (ret) {
@@ -569,7 +578,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,int id, ret;if (!mops || !mops->write_complete || !mops->state || - !mops->write_init || (!mops->write && !mops->write_sg) || + (!mops->write && !mops->write_sg) || (mops->write && mops->write_sg)) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );return NULL ;
Signature · 1 line
2.26.3
Moritz Fischer
maintainerRe: [PATCH v4 1/7] fpga-mgr: wrap the write_init() op
Quoted 17 lines · On Fri, Jun 25, 2021 at 12:51:42PM -0700, [email protected] wrote:
From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a
write_init()op if there is nothing for it do. So add a wrapper and move the op checking. Default to success.Signed-off-by Tom Rix <[email protected]>--- drivers/fpga/fpga-mgr.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-)drivers/fpga/fpga-mgr.c +0
-0
drivers/fpga/fpga-mgr.c struct fpga_mgr_devres { · line 32
Quoted 8 lines
drivers/fpga/fpga-mgr.c +5
-0@@ -25,6 +25,15 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline int fpga_mgr_write_init (struct fpga_manager *mgr, +struct fpga_image_info *info, +const char *buf,size_t count) +{ +if (mgr->mops->write_init)
Will you need a if (mgr->mops && mgr->mops->write_init) here later?
Quoted 31 lines
drivers/fpga/fpga-mgr.c +7
-3+return mgr->mops->write_init (mgr, info, buf, count); +return 0 ; +} +/** * fpga_image_info_alloc - Allocate an FPGA image info struct * @dev: owning device @@ -83,9 +92,9 @@ static int fpga_mgr_write_init_buf(struct fpga_manager *mgr, mgr->state =FPGA_MGR_STATE_WRITE_INIT ;if (!mgr->mops->initial_header_size) - ret = mgr->mops->write_init (mgr, info,NULL ,0 ); + ret =fpga_mgr_write_init (mgr, info,NULL ,0 );else - ret = mgr->mops->write_init ( + ret =fpga_mgr_write_init ( mgr, info, buf,min (mgr->mops->initial_header_size, count));if (ret) {@@ -569,7 +578,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,int id, ret;if (!mops || !mops->write_complete || !mops->state || - !mops->write_init || (!mops->write && !mops->write_sg) || + (!mops->write && !mops->write_sg) || (mops->write && mops->write_sg)) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );return NULL ;Signature · 1 line
2.26.3
Looks good to me, I might reword the commit message some when applying.
- Moritz
Tom Rix
reviewerRe: [PATCH v4 1/7] fpga-mgr: wrap the write_init() op
Quoted 21 lines · Moritz Fischer
Quoted 20 lines · On Fri, Jun 25, 2021 at 12:51:42PM -0700, [email protected] wrote:
From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a
write_init()op if there is nothing for it do. So add a wrapper and move the op checking. Default to success.Signed-off-by Tom Rix <[email protected]>--- drivers/fpga/fpga-mgr.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-)drivers/fpga/fpga-mgr.c +0
-0@@ -25,6 +25,15 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; };
Quoted 6 lines
drivers/fpga/fpga-mgr.c struct fpga_mgr_devres { · line 32
Quoted 5 lines
drivers/fpga/fpga-mgr.c +5
-0+static inline int fpga_mgr_write_init (struct fpga_manager *mgr, +struct fpga_image_info *info, +const char *buf,size_t count) +{ +if (mgr->mops->write_init)Will you need a if (
mgr->mops&&mgr->mops->write_init) here later?
This was changed from v3 based on Yilun's comment
https://lore.kernel.org/linux-fpga/20210624075414.GA44700@yilunxu-OptiPlex-7050/
This is checked on creation
Quoted 22 lines
Quoted 22 lines
drivers/fpga/fpga-mgr.c +6
-2+return mgr->mops->write_init (mgr, info, buf, count); +return 0 ; +} +/** * fpga_image_info_alloc - Allocate an FPGA image info struct * @dev: owning device @@ -83,9 +92,9 @@ static int fpga_mgr_write_init_buf(struct fpga_manager *mgr, mgr->state =FPGA_MGR_STATE_WRITE_INIT ;if (!mgr->mops->initial_header_size) - ret = mgr->mops->write_init (mgr, info,NULL ,0 ); + ret =fpga_mgr_write_init (mgr, info,NULL ,0 );else - ret = mgr->mops->write_init ( + ret =fpga_mgr_write_init ( mgr, info, buf,min (mgr->mops->initial_header_size, count));if (ret) {@@ -569,7 +578,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,int id, ret;
Quoted 10 lines
Quoted 9 lines
drivers/fpga/fpga-mgr.c +1
-1if (!mops || !mops->write_complete || !mops->state || - !mops->write_init || (!mops->write && !mops->write_sg) || + (!mops->write && !mops->write_sg) || (mops->write && mops->write_sg)) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );return NULL ;Signature · 1 line
2.26.3Looks good to me, I might reword the commit message some when applying.
That is fine.
Thanks
Tom
Quoted 2 lines
- Moritz
[email protected]
reviewerPATCH v4 2/7 fpga-mgr: make write_complete() op optional
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a write_complete function if there is nothing. Move the op check to the existing wrapper. Default to success.
Remove noop function.
--- drivers/fpga/fpga-mgr.c | 45 +++++++++++++++++++------------------- drivers/fpga/zynqmp-fpga.c | 7 ------ 2 files changed, 23 insertions(+), 29 deletions(-)
drivers/fpga/fpga-mgr.c +23 -22
@@ -25,6 +25,28 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +/* +* After all the FPGA image has been written, do the device specific steps to +* finish and set the FPGA into operating mode. +*/ +static inline int fpga_mgr_write_complete (struct fpga_manager *mgr, +struct fpga_image_info *info) +{ +int ret =0 ; + + mgr->state =FPGA_MGR_STATE_WRITE_COMPLETE ; +if (mgr->mops->write_complete) + ret = mgr->mops->write_complete (mgr, info); +if (ret) { +dev_err (&mgr->dev,"Error after writing image data to FPGA\n" ); + mgr->state =FPGA_MGR_STATE_WRITE_COMPLETE_ERR ; +return ret; + } + mgr->state =FPGA_MGR_STATE_OPERATING ; + +return 0 ; +} +static inline int fpga_mgr_write_init (struct fpga_manager *mgr,struct fpga_image_info *info,const char *buf,size_t count)
@@ -146,27 +168,6 @@ static int fpga_mgr_write_init_sg(struct fpga_manager *mgr,return ret; } -/* -* After all the FPGA image has been written, do the device specific steps to -* finish and set the FPGA into operating mode. -*/ -static int fpga_mgr_write_complete (struct fpga_manager *mgr, -struct fpga_image_info *info) -{ -int ret; - - mgr->state =FPGA_MGR_STATE_WRITE_COMPLETE ; - ret = mgr->mops->write_complete (mgr, info); -if (ret) { -dev_err (&mgr->dev,"Error after writing image data to FPGA\n" ); - mgr->state =FPGA_MGR_STATE_WRITE_COMPLETE_ERR ; -return ret; - } - mgr->state =FPGA_MGR_STATE_OPERATING ; - -return 0 ; -} -/** * fpga_mgr_buf_load_sg - load fpga from image in buffer from a scatter list * @mgr: fpga manager
@@ -577,7 +578,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,struct fpga_manager *mgr;int id, ret; -if (!mops || !mops->write_complete || !mops->state || +if (!mops || !mops->state || (!mops->write && !mops->write_sg) || (mops->write && mops->write_sg)) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );
drivers/fpga/zynqmp-fpga.c +0 -7
@@ -66,12 +66,6 @@ static int zynqmp_fpga_ops_write(struct fpga_manager *mgr,return ret; } -static int zynqmp_fpga_ops_write_complete (struct fpga_manager *mgr, -struct fpga_image_info *info) -{ -return 0 ; -} -static enum fpga_mgr_states zynqmp_fpga_ops_state (struct fpga_manager *mgr) {u32 status =0 ;
@@ -87,7 +81,6 @@ static const struct fpga_manager_ops zynqmp_fpga_ops = { .state = zynqmp_fpga_ops_state, .write_init = zynqmp_fpga_ops_write_init, .write = zynqmp_fpga_ops_write, - .write_complete = zynqmp_fpga_ops_write_complete, };static int zynqmp_fpga_probe (struct platform_device *pdev)
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 3/7 fpga-mgr: wrap the write() op
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a write function. Move the op check to the wrapper. Default to -EOPNOTSUP so its users will fail gracefully.
--- drivers/fpga/fpga-mgr.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-)
drivers/fpga/fpga-mgr.c +10 -5
@@ -25,6 +25,13 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline int fpga_mgr_write (struct fpga_manager *mgr,const char *buf,size_t count) +{ +if (mgr->mops->write) +return mgr->mops->write (mgr, buf, count); +return -EOPNOTSUPP ; +} +/* * After all the FPGA image has been written, do the device specific steps to * finish and set the FPGA into operating mode.
@@ -204,7 +211,7 @@ static int fpga_mgr_buf_load_sg(struct fpga_manager *mgr,sg_miter_start (&miter, sgt->sgl, sgt->nents,SG_MITER_FROM_SG );while (sg_miter_next (&miter)) { - ret = mgr->mops->write (mgr, miter.addr, miter.length); + ret =fpga_mgr_write (mgr, miter.addr, miter.length);if (ret)break ; }
@@ -234,7 +241,7 @@ static int fpga_mgr_buf_load_mapped(struct fpga_manager *mgr,* Write the FPGA image to the FPGA. */ mgr->state =FPGA_MGR_STATE_WRITE ; - ret = mgr->mops->write (mgr, buf, count); + ret =fpga_mgr_write (mgr, buf, count);if (ret) {dev_err (&mgr->dev,"Error while writing image data to FPGA\n" ); mgr->state =FPGA_MGR_STATE_WRITE_ERR ;
@@ -578,9 +585,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,struct fpga_manager *mgr;int id, ret; -if (!mops || !mops->state || - (!mops->write && !mops->write_sg) || - (mops->write && mops->write_sg)) { +if (!mops || !mops->state) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );return NULL ; }
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 4/7 fpga-mgr: wrap the status() op
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager is not required to provide a status() op. Add a wrapper consistent with the other op wrappers. Move the op check to the wrapper.
Default to 0, no errors to report.
--- drivers/fpga/fpga-mgr.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/fpga/fpga-mgr.c +8 -4
@@ -25,6 +25,13 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline u64 fpga_mgr_status (struct fpga_manager *mgr) +{ +if (mgr->mops->status) +return mgr->mops->status (mgr); +return 0 ; +} +static inline int fpga_mgr_write (struct fpga_manager *mgr,const char *buf,size_t count) {if (mgr->mops->write)
@@ -434,10 +441,7 @@ static ssize_t status_show(struct device *dev,u64 status;int len =0 ; -if (!mgr->mops->status) -return -ENOENT ; - - status = mgr->mops->status (mgr); + status =fpga_mgr_status (mgr);if (status &FPGA_MGR_STATUS_OPERATION_ERR ) len +=sprintf (buf + len,"reconfig operation error\n" );
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 5/7 fpga-mgr: wrap the state() op
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a state() op. Add a wrapper consistent with the other op wrappers. Move op check to wrapper.
Default to FPGA_MGR_STATE_UNKNOWN, what noop state() ops use. Remove unneeded noop state() ops
--- drivers/fpga/dfl-fme-mgr.c | 6 ------ drivers/fpga/fpga-mgr.c | 11 +++++++++-- drivers/fpga/stratix10-soc.c | 6 ------ drivers/fpga/ts73xx-fpga.c | 6 ------ 4 files changed, 9 insertions(+), 20 deletions(-)
drivers/fpga/dfl-fme-mgr.c +0 -6
@@ -252,11 +252,6 @@ static int fme_mgr_write_complete(struct fpga_manager *mgr,return 0 ; } -static enum fpga_mgr_states fme_mgr_state (struct fpga_manager *mgr) -{ -return FPGA_MGR_STATE_UNKNOWN ; -} -static u64 fme_mgr_status (struct fpga_manager *mgr) {struct fme_mgr_priv *priv = mgr->priv;
@@ -268,7 +263,6 @@ static const struct fpga_manager_ops fme_mgr_ops = { .write_init = fme_mgr_write_init, .write = fme_mgr_write, .write_complete = fme_mgr_write_complete, - .state = fme_mgr_state, .status = fme_mgr_status, };
drivers/fpga/fpga-mgr.c +9 -2
@@ -25,6 +25,13 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline enum fpga_mgr_states fpga_mgr_state (struct fpga_manager *mgr) +{ +if (mgr->mops->state) +return mgr->mops->state (mgr); +return FPGA_MGR_STATE_UNKNOWN ; +} +static inline u64 fpga_mgr_status (struct fpga_manager *mgr) {if (mgr->mops->status)
@@ -589,7 +596,7 @@ struct fpga_manager *fpga_mgr_create(struct device *parent, const char *name,struct fpga_manager *mgr;int id, ret; -if (!mops || !mops->state) { +if (!mops) {dev_err (parent,"Attempt to register without fpga_manager_ops\n" );return NULL ; }
@@ -707,7 +714,7 @@ int fpga_mgr_register(struct fpga_manager *mgr)* from device. FPGA may be in reset mode or may have been programmed * by bootloader or EEPROM. */ - mgr->state = mgr->mops->state (mgr); + mgr->state =fpga_mgr_state (mgr); ret =device_add (&mgr->dev);if (ret)
drivers/fpga/stratix10-soc.c +0 -6
@@ -388,13 +388,7 @@ static int s10_ops_write_complete(struct fpga_manager *mgr,return ret; } -static enum fpga_mgr_states s10_ops_state (struct fpga_manager *mgr) -{ -return FPGA_MGR_STATE_UNKNOWN ; -} -static const struct fpga_manager_ops s10_ops = { - .state = s10_ops_state, .write_init = s10_ops_write_init, .write = s10_ops_write, .write_complete = s10_ops_write_complete,
drivers/fpga/ts73xx-fpga.c +0 -6
@@ -32,11 +32,6 @@ struct ts73xx_fpga_priv {struct device *dev; }; -static enum fpga_mgr_states ts73xx_fpga_state (struct fpga_manager *mgr) -{ -return FPGA_MGR_STATE_UNKNOWN ; -} -static int ts73xx_fpga_write_init (struct fpga_manager *mgr,struct fpga_image_info *info,const char *buf,size_t count)
@@ -98,7 +93,6 @@ static int ts73xx_fpga_write_complete(struct fpga_manager *mgr, }static const struct fpga_manager_ops ts73xx_fpga_ops = { - .state = ts73xx_fpga_state, .write_init = ts73xx_fpga_write_init, .write = ts73xx_fpga_write, .write_complete = ts73xx_fpga_write_complete,
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 6/7 fpga-mgr: wrap the fpga_remove() op
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager is not required to provide a fpga_remove() op. Add a wrapper consistent with the other op wrappers. Move op check to wrapper.
--- drivers/fpga/fpga-mgr.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/fpga/fpga-mgr.c +7 -2
@@ -25,6 +25,12 @@ struct fpga_mgr_devres {struct fpga_manager *mgr; }; +static inline void fpga_mgr_fpga_remove (struct fpga_manager *mgr) +{ +if (mgr->mops->fpga_remove) + mgr->mops->fpga_remove (mgr); +} +static inline enum fpga_mgr_states fpga_mgr_state (struct fpga_manager *mgr) {if (mgr->mops->state)
@@ -745,8 +751,7 @@ void fpga_mgr_unregister(struct fpga_manager *mgr)* If the low level driver provides a method for putting fpga into * a desired state upon unregister, do it. */ -if (mgr->mops->fpga_remove) - mgr->mops->fpga_remove (mgr); +fpga_mgr_fpga_remove (mgr);device_unregister (&mgr->dev); }
Signature · 1 line
2.26.3
[email protected]
reviewerPATCH v4 7/7 fpga-mgr: wrap the write_sg() op
In reply to [email protected]From: Tom Rix <[email protected]>
An FPGA manager should not be required to provide a write_sg function. Move the op check to the wrapper. Default to -EOPNOTSUP so its users will fail gracefully.
--- drivers/fpga/fpga-mgr.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
drivers/fpga/fpga-mgr.c +9 -1
@@ -83,6 +83,14 @@ static inline int fpga_mgr_write_init(struct fpga_manager *mgr,return 0 ; } +static inline int fpga_mgr_write_sg (struct fpga_manager *mgr, +struct sg_table *sgt) +{ +if (mgr->mops->write_sg) +return mgr->mops->write_sg (mgr, sgt); +return -EOPNOTSUPP ; +} +/** * fpga_image_info_alloc - Allocate an FPGA image info struct * @dev: owning device
@@ -225,7 +233,7 @@ static int fpga_mgr_buf_load_sg(struct fpga_manager *mgr,/* Write the FPGA image to the FPGA. */ mgr->state =FPGA_MGR_STATE_WRITE ;if (mgr->mops->write_sg) { - ret = mgr->mops->write_sg (mgr, sgt); + ret =fpga_mgr_write_sg (mgr, sgt); }else {struct sg_mapping_iter miter;
Signature · 1 line
2.26.3
Moritz Fischer
maintainerQuoted 32 lines · On Fri, Jun 25, 2021 at 12:51:40PM -0700, [email protected] wrote:
From: Tom Rix <[email protected]>
As followup from
https://lore.kernel.org/linux-fpga/[email protected]/Boards should not be required to have noop functions. So improve or create fpga-mgr wrappers for the
fpga_manager_ops. Remove the noop functions.
Refactor fpga-mgr to use the wrappers.Changes from
v1:
commit subject,logv2:
rebase to next-20210623v3:
remove mops check
addwrite_sgwrapper
drop 'fpga-mgr: collect wappers and change to inline'Tom Rix (7):
fpga-mgr: wrap thewrite_init()op
fpga-mgr: makewrite_complete()op optional
fpga-mgr: wrap thewrite()op
fpga-mgr: wrap thestatus()op
fpga-mgr: wrap thestate()op
fpga-mgr: wrap thefpga_remove()op
fpga-mgr: wrap thewrite_sg()op
Quoted 10 lines
drivers/fpga/dfl-fme-mgr.c | 6 -- drivers/fpga/fpga-mgr.c | 111 +++++++++++++++++++++++------------ drivers/fpga/stratix10-soc.c | 6 -- drivers/fpga/ts73xx-fpga.c | 6 -- drivers/fpga/zynqmp-fpga.c | 7 --- 5 files changed, 75 insertions(+), 61 deletions(-)Signature · 1 line
2.26.3
Appled to for-next,
Thanks