linux/drivers/of/base.c

2056 lines
52 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0+
/*
* Procedures for creating, accessing and interpreting the device tree.
*
* Paul Mackerras August 1996.
* Copyright (C) 1996-2005 Paul Mackerras.
*
* Adapted for 64bit PowerPC by Dave Engebretsen and Peter Bergner.
* {engebret|bergner}@us.ibm.com
*
* Adapted for sparc and sparc64 by David S. Miller davem@davemloft.net
*
* Reconsolidated from arch/x/kernel/prom.c by Stephen Rothwell and
* Grant Likely.
*/
#define pr_fmt(fmt) "OF: " fmt
#include <linux/console.h>
#include <linux/ctype.h>
#include <linux/cpu.h>
#include <linux/module.h>
#include <linux/of.h>
#include <linux/of_device.h>
#include <linux/of_graph.h>
#include <linux/spinlock.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/proc_fs.h>
#include "of_private.h"
LIST_HEAD(aliases_lookup);
struct device_node *of_root;
EXPORT_SYMBOL(of_root);
struct device_node *of_chosen;
fbdev: simplefb: fix Kconfig dependencies Configurations with both CONFIG_FB_SIMPLE=y and CONFIG_DRM_SIMPLEDRM=m are allowed by Kconfig because the 'depends on !DRM_SIMPLEDRM' dependency does not disallow FB_SIMPLE as long as SIMPLEDRM is not built-in. This can however result in a build failure when cfb_fillrect() etc are then also in loadable modules: x86_64-linux-ld: drivers/video/fbdev/simplefb.o:(.rodata+0x1f8): undefined reference to `cfb_fillrect' x86_64-linux-ld: drivers/video/fbdev/simplefb.o:(.rodata+0x200): undefined reference to `cfb_copyarea' x86_64-linux-ld: drivers/video/fbdev/simplefb.o:(.rodata+0x208): undefined reference to `cfb_imageblit' To work around this, change FB_SIMPLE to be a 'tristate' symbol, which still allows both to be =m together, but not one of them to be =y if the other one is =m. If a distro kernel picks this configuration, it can be determined by local policy which of the two modules gets loaded. The 'of_chosen' export is needed as this is the first loadable module referencing it. Alternatively, the Kconfig dependency could be changed to 'depends on DRM_SIMPLEDRM=n', which would forbid the configuration with both drivers. Fixes: 11e8f5fd223b ("drm: Add simpledrm driver") Acked-by: Rob Herring <robh@kernel.org> # for drivers/of/ Link: https://lore.kernel.org/all/20210721151839.2484245-1-arnd@kernel.org/ Signed-off-by: Arnd Bergmann <arnd@arndb.de> Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> # fbdev support Cc: Maxime Ripard <maxime@cerno.tech> Cc: Liam Girdwood <lgirdwood@gmail.com> Cc: Mark Brown <broonie@kernel.org> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Borislav Petkov <bp@suse.de> Cc: Javier Martinez Canillas <javierm@redhat.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Geert Uytterhoeven <geert+renesas@glider.be> Cc: Peter Collingbourne <pcc@google.com> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: dri-devel@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Cc: <stable@vger.kernel.org> # v5.14+ Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de> Link: https://patchwork.freedesktop.org/patch/msgid/20210928145243.1098064-1-arnd@kernel.org Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
2021-09-28 14:52:10 +00:00
EXPORT_SYMBOL(of_chosen);
struct device_node *of_aliases;
struct device_node *of_stdout;
static const char *of_stdout_options;
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
struct kset *of_kset;
/*
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
* Used to protect the of_aliases, to hold off addition of nodes to sysfs.
* This mutex must be held whenever modifications are being made to the
* device tree. The of_{attach,detach}_node() and
* of_{add,remove,update}_property() helpers make sure this happens.
*/
DEFINE_MUTEX(of_mutex);
/* use when traversing tree through the child, sibling,
* or parent members of struct device_node.
*/
DEFINE_RAW_SPINLOCK(devtree_lock);
bool of_node_name_eq(const struct device_node *np, const char *name)
{
const char *node_name;
size_t len;
if (!np)
return false;
node_name = kbasename(np->full_name);
len = strchrnul(node_name, '@') - node_name;
return (strlen(name) == len) && (strncmp(node_name, name, len) == 0);
}
EXPORT_SYMBOL(of_node_name_eq);
bool of_node_name_prefix(const struct device_node *np, const char *prefix)
{
if (!np)
return false;
return strncmp(kbasename(np->full_name), prefix, strlen(prefix)) == 0;
}
EXPORT_SYMBOL(of_node_name_prefix);
static bool __of_node_is_type(const struct device_node *np, const char *type)
{
const char *match = __of_get_property(np, "device_type", NULL);
return np && match && type && !strcmp(match, type);
}
int of_bus_n_addr_cells(struct device_node *np)
{
u32 cells;
for (; np; np = np->parent)
if (!of_property_read_u32(np, "#address-cells", &cells))
return cells;
/* No #address-cells property for the root node */
return OF_ROOT_NODE_ADDR_CELLS_DEFAULT;
}
int of_n_addr_cells(struct device_node *np)
{
if (np->parent)
np = np->parent;
return of_bus_n_addr_cells(np);
}
EXPORT_SYMBOL(of_n_addr_cells);
int of_bus_n_size_cells(struct device_node *np)
{
u32 cells;
for (; np; np = np->parent)
if (!of_property_read_u32(np, "#size-cells", &cells))
return cells;
/* No #size-cells property for the root node */
return OF_ROOT_NODE_SIZE_CELLS_DEFAULT;
}
int of_n_size_cells(struct device_node *np)
{
if (np->parent)
np = np->parent;
return of_bus_n_size_cells(np);
}
EXPORT_SYMBOL(of_n_size_cells);
#ifdef CONFIG_NUMA
int __weak of_node_to_nid(struct device_node *np)
{
return NUMA_NO_NODE;
}
#endif
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
#define OF_PHANDLE_CACHE_BITS 7
#define OF_PHANDLE_CACHE_SZ BIT(OF_PHANDLE_CACHE_BITS)
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
static struct device_node *phandle_cache[OF_PHANDLE_CACHE_SZ];
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
static u32 of_phandle_cache_hash(phandle handle)
{
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
return hash_32(handle, OF_PHANDLE_CACHE_BITS);
}
/*
* Caller must hold devtree_lock.
*/
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
void __of_phandle_cache_inv_entry(phandle handle)
{
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
u32 handle_hash;
struct device_node *np;
if (!handle)
return;
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
handle_hash = of_phandle_cache_hash(handle);
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
np = phandle_cache[handle_hash];
if (np && handle == np->phandle)
phandle_cache[handle_hash] = NULL;
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
}
void __init of_core_init(void)
{
struct device_node *np;
of_platform_register_reconfig_notifier();
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
/* Create the kset, and register existing nodes */
mutex_lock(&of_mutex);
of_kset = kset_create_and_add("devicetree", NULL, firmware_kobj);
if (!of_kset) {
mutex_unlock(&of_mutex);
pr_err("failed to register existing nodes\n");
return;
}
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
for_each_of_allnodes(np) {
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
__of_attach_node_sysfs(np);
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
if (np->phandle && !phandle_cache[of_phandle_cache_hash(np->phandle)])
phandle_cache[of_phandle_cache_hash(np->phandle)] = np;
}
mutex_unlock(&of_mutex);
/* Symlink in /proc as required by userspace ABI */
if (of_root)
proc_symlink("device-tree", NULL, "/sys/firmware/devicetree/base");
}
static struct property *__of_find_property(const struct device_node *np,
const char *name, int *lenp)
{
struct property *pp;
if (!np)
return NULL;
for (pp = np->properties; pp; pp = pp->next) {
if (of_prop_cmp(pp->name, name) == 0) {
if (lenp)
*lenp = pp->length;
break;
}
}
return pp;
}
struct property *of_find_property(const struct device_node *np,
const char *name,
int *lenp)
{
struct property *pp;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
pp = __of_find_property(np, name, lenp);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return pp;
}
EXPORT_SYMBOL(of_find_property);
struct device_node *__of_find_all_nodes(struct device_node *prev)
{
struct device_node *np;
if (!prev) {
np = of_root;
} else if (prev->child) {
np = prev->child;
} else {
/* Walk back up looking for a sibling, or the end of the structure */
np = prev;
while (np->parent && !np->sibling)
np = np->parent;
np = np->sibling; /* Might be null at the end of the tree */
}
return np;
}
/**
* of_find_all_nodes - Get next node in global list
* @prev: Previous node or NULL to start iteration
* of_node_put() will be called on it
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_all_nodes(struct device_node *prev)
{
struct device_node *np;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
np = __of_find_all_nodes(prev);
of_node_get(np);
of_node_put(prev);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_all_nodes);
/*
* Find a property with a given name for a given node
* and return the value.
*/
const void *__of_get_property(const struct device_node *np,
const char *name, int *lenp)
{
struct property *pp = __of_find_property(np, name, lenp);
return pp ? pp->value : NULL;
}
/*
* Find a property with a given name for a given node
* and return the value.
*/
const void *of_get_property(const struct device_node *np, const char *name,
int *lenp)
{
struct property *pp = of_find_property(np, name, lenp);
return pp ? pp->value : NULL;
}
EXPORT_SYMBOL(of_get_property);
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
/**
* __of_device_is_compatible() - Check if the node matches given constraints
* @device: pointer to node
* @compat: required compatible string, NULL or "" for any match
* @type: required device_type value, NULL or "" for any match
* @name: required node name, NULL or "" for any match
*
* Checks if the given @compat, @type and @name strings match the
* properties of the given @device. A constraints can be skipped by
* passing NULL or an empty string as the constraint.
*
* Returns 0 for no match, and a positive integer on match. The return
* value is a relative score with larger values indicating better
* matches. The score is weighted for the most specific compatible value
* to get the highest score. Matching type is next, followed by matching
* name. Practically speaking, this results in the following priority
* order for matches:
*
* 1. specific compatible && type && name
* 2. specific compatible && type
* 3. specific compatible && name
* 4. specific compatible
* 5. general compatible && type && name
* 6. general compatible && type
* 7. general compatible && name
* 8. general compatible
* 9. type && name
* 10. type
* 11. name
*/
static int __of_device_is_compatible(const struct device_node *device,
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
const char *compat, const char *type, const char *name)
{
struct property *prop;
const char *cp;
int index = 0, score = 0;
/* Compatible match has highest priority */
if (compat && compat[0]) {
prop = __of_find_property(device, "compatible", NULL);
for (cp = of_prop_next_string(prop, NULL); cp;
cp = of_prop_next_string(prop, cp), index++) {
if (of_compat_cmp(cp, compat, strlen(compat)) == 0) {
score = INT_MAX/2 - (index << 2);
break;
}
}
if (!score)
return 0;
}
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
/* Matching type is better than matching name */
if (type && type[0]) {
if (!__of_node_is_type(device, type))
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
return 0;
score += 2;
}
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
/* Matching name is a bit better than not */
if (name && name[0]) {
if (!of_node_name_eq(device, name))
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
return 0;
score++;
}
return score;
}
/** Checks if the given "compat" string matches one of the strings in
* the device's "compatible" property
*/
int of_device_is_compatible(const struct device_node *device,
const char *compat)
{
unsigned long flags;
int res;
raw_spin_lock_irqsave(&devtree_lock, flags);
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
res = __of_device_is_compatible(device, compat, NULL, NULL);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return res;
}
EXPORT_SYMBOL(of_device_is_compatible);
/** Checks if the device is compatible with any of the entries in
* a NULL terminated array of strings. Returns the best match
* score or 0.
*/
int of_device_compatible_match(const struct device_node *device,
const char *const *compat)
{
unsigned int tmp, score = 0;
if (!compat)
return 0;
while (*compat) {
tmp = of_device_is_compatible(device, *compat);
if (tmp > score)
score = tmp;
compat++;
}
return score;
}
EXPORT_SYMBOL_GPL(of_device_compatible_match);
/**
* of_machine_is_compatible - Test root of device tree for a given compatible value
* @compat: compatible string to look for in root node's compatible property.
*
* Return: A positive integer if the root node has the given value in its
* compatible property.
*/
int of_machine_is_compatible(const char *compat)
{
struct device_node *root;
int rc = 0;
root = of_find_node_by_path("/");
if (root) {
rc = of_device_is_compatible(root, compat);
of_node_put(root);
}
return rc;
}
EXPORT_SYMBOL(of_machine_is_compatible);
static bool __of_device_is_status(const struct device_node *device,
const char * const*strings)
{
const char *status;
int statlen;
if (!device)
return false;
status = __of_get_property(device, "status", &statlen);
if (status == NULL)
return false;
if (statlen > 0) {
while (*strings) {
unsigned int len = strlen(*strings);
if ((*strings)[len - 1] == '-') {
if (!strncmp(status, *strings, len))
return true;
} else {
if (!strcmp(status, *strings))
return true;
}
strings++;
}
}
return false;
}
/**
* __of_device_is_available - check if a device is available for use
*
* @device: Node to check for availability, with locks already held
*
* Return: True if the status property is absent or set to "okay" or "ok",
* false otherwise
*/
static bool __of_device_is_available(const struct device_node *device)
{
static const char * const ok[] = {"okay", "ok", NULL};
if (!device)
return false;
return !__of_get_property(device, "status", NULL) ||
__of_device_is_status(device, ok);
}
/**
* of_device_is_available - check if a device is available for use
*
* @device: Node to check for availability
*
* Return: True if the status property is absent or set to "okay" or "ok",
* false otherwise
*/
bool of_device_is_available(const struct device_node *device)
{
unsigned long flags;
bool res;
raw_spin_lock_irqsave(&devtree_lock, flags);
res = __of_device_is_available(device);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return res;
}
EXPORT_SYMBOL(of_device_is_available);
/**
* __of_device_is_fail - check if a device has status "fail" or "fail-..."
*
* @device: Node to check status for, with locks already held
*
* Return: True if the status property is set to "fail" or "fail-..." (for any
* error code suffix), false otherwise
*/
static bool __of_device_is_fail(const struct device_node *device)
{
static const char * const fail[] = {"fail", "fail-", NULL};
return __of_device_is_status(device, fail);
}
/**
* of_device_is_big_endian - check if a device has BE registers
*
* @device: Node to check for endianness
*
* Return: True if the device has a "big-endian" property, or if the kernel
* was compiled for BE *and* the device has a "native-endian" property.
* Returns false otherwise.
*
* Callers would nominally use ioread32be/iowrite32be if
* of_device_is_big_endian() == true, or readl/writel otherwise.
*/
bool of_device_is_big_endian(const struct device_node *device)
{
if (of_property_read_bool(device, "big-endian"))
return true;
if (IS_ENABLED(CONFIG_CPU_BIG_ENDIAN) &&
of_property_read_bool(device, "native-endian"))
return true;
return false;
}
EXPORT_SYMBOL(of_device_is_big_endian);
/**
* of_get_parent - Get a node's parent if any
* @node: Node to get parent
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_get_parent(const struct device_node *node)
{
struct device_node *np;
unsigned long flags;
if (!node)
return NULL;
raw_spin_lock_irqsave(&devtree_lock, flags);
np = of_node_get(node->parent);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_get_parent);
/**
* of_get_next_parent - Iterate to a node's parent
* @node: Node to get parent of
*
* This is like of_get_parent() except that it drops the
* refcount on the passed node, making it suitable for iterating
* through a node's parents.
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_get_next_parent(struct device_node *node)
{
struct device_node *parent;
unsigned long flags;
if (!node)
return NULL;
raw_spin_lock_irqsave(&devtree_lock, flags);
parent = of_node_get(node->parent);
of_node_put(node);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return parent;
}
EXPORT_SYMBOL(of_get_next_parent);
static struct device_node *__of_get_next_child(const struct device_node *node,
struct device_node *prev)
{
struct device_node *next;
if (!node)
return NULL;
next = prev ? prev->sibling : node->child;
of_node_get(next);
of_node_put(prev);
return next;
}
#define __for_each_child_of_node(parent, child) \
for (child = __of_get_next_child(parent, NULL); child != NULL; \
child = __of_get_next_child(parent, child))
/**
* of_get_next_child - Iterate a node childs
* @node: parent node
* @prev: previous child of the parent node, or NULL to get first
*
* Return: A node pointer with refcount incremented, use of_node_put() on
* it when done. Returns NULL when prev is the last child. Decrements the
* refcount of prev.
*/
struct device_node *of_get_next_child(const struct device_node *node,
struct device_node *prev)
{
struct device_node *next;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
next = __of_get_next_child(node, prev);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return next;
}
EXPORT_SYMBOL(of_get_next_child);
/**
* of_get_next_available_child - Find the next available child node
* @node: parent node
* @prev: previous child of the parent node, or NULL to get first
*
* This function is like of_get_next_child(), except that it
* automatically skips any disabled nodes (i.e. status = "disabled").
*/
struct device_node *of_get_next_available_child(const struct device_node *node,
struct device_node *prev)
{
struct device_node *next;
unsigned long flags;
if (!node)
return NULL;
raw_spin_lock_irqsave(&devtree_lock, flags);
next = prev ? prev->sibling : node->child;
for (; next; next = next->sibling) {
if (!__of_device_is_available(next))
continue;
if (of_node_get(next))
break;
}
of_node_put(prev);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return next;
}
EXPORT_SYMBOL(of_get_next_available_child);
/**
* of_get_next_cpu_node - Iterate on cpu nodes
* @prev: previous child of the /cpus node, or NULL to get first
*
* Unusable CPUs (those with the status property set to "fail" or "fail-...")
* will be skipped.
*
* Return: A cpu node pointer with refcount incremented, use of_node_put()
* on it when done. Returns NULL when prev is the last child. Decrements
* the refcount of prev.
*/
struct device_node *of_get_next_cpu_node(struct device_node *prev)
{
struct device_node *next = NULL;
unsigned long flags;
struct device_node *node;
if (!prev)
node = of_find_node_by_path("/cpus");
raw_spin_lock_irqsave(&devtree_lock, flags);
if (prev)
next = prev->sibling;
else if (node) {
next = node->child;
of_node_put(node);
}
for (; next; next = next->sibling) {
if (__of_device_is_fail(next))
continue;
if (!(of_node_name_eq(next, "cpu") ||
__of_node_is_type(next, "cpu")))
continue;
if (of_node_get(next))
break;
}
of_node_put(prev);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return next;
}
EXPORT_SYMBOL(of_get_next_cpu_node);
/**
* of_get_compatible_child - Find compatible child node
* @parent: parent node
* @compatible: compatible string
*
* Lookup child node whose compatible property contains the given compatible
* string.
*
* Return: a node pointer with refcount incremented, use of_node_put() on it
* when done; or NULL if not found.
*/
struct device_node *of_get_compatible_child(const struct device_node *parent,
const char *compatible)
{
struct device_node *child;
for_each_child_of_node(parent, child) {
if (of_device_is_compatible(child, compatible))
break;
}
return child;
}
EXPORT_SYMBOL(of_get_compatible_child);
/**
* of_get_child_by_name - Find the child node by name for a given parent
* @node: parent node
* @name: child name to look for.
*
* This function looks for child node for given matching name
*
* Return: A node pointer if found, with refcount incremented, use
* of_node_put() on it when done.
* Returns NULL if node is not found.
*/
struct device_node *of_get_child_by_name(const struct device_node *node,
const char *name)
{
struct device_node *child;
for_each_child_of_node(node, child)
if (of_node_name_eq(child, name))
break;
return child;
}
EXPORT_SYMBOL(of_get_child_by_name);
struct device_node *__of_find_node_by_path(struct device_node *parent,
const char *path)
{
struct device_node *child;
int len;
len = strcspn(path, "/:");
if (!len)
return NULL;
__for_each_child_of_node(parent, child) {
const char *name = kbasename(child->full_name);
if (strncmp(path, name, len) == 0 && (strlen(name) == len))
return child;
}
return NULL;
}
struct device_node *__of_find_node_by_full_path(struct device_node *node,
const char *path)
{
const char *separator = strchr(path, ':');
while (node && *path == '/') {
struct device_node *tmp = node;
path++; /* Increment past '/' delimiter */
node = __of_find_node_by_path(node, path);
of_node_put(tmp);
path = strchrnul(path, '/');
if (separator && separator < path)
break;
}
return node;
}
/**
* of_find_node_opts_by_path - Find a node matching a full OF path
* @path: Either the full path to match, or if the path does not
* start with '/', the name of a property of the /aliases
* node (an alias). In the case of an alias, the node
* matching the alias' value will be returned.
* @opts: Address of a pointer into which to store the start of
* an options string appended to the end of the path with
* a ':' separator.
*
* Valid paths:
* * /foo/bar Full path
* * foo Valid alias
* * foo/bar Valid alias + relative path
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_node_opts_by_path(const char *path, const char **opts)
{
struct device_node *np = NULL;
struct property *pp;
unsigned long flags;
const char *separator = strchr(path, ':');
if (opts)
*opts = separator ? separator + 1 : NULL;
if (strcmp(path, "/") == 0)
return of_node_get(of_root);
/* The path could begin with an alias */
if (*path != '/') {
int len;
const char *p = separator;
if (!p)
p = strchrnul(path, '/');
len = p - path;
/* of_aliases must not be NULL */
if (!of_aliases)
return NULL;
for_each_property_of_node(of_aliases, pp) {
if (strlen(pp->name) == len && !strncmp(pp->name, path, len)) {
np = of_find_node_by_path(pp->value);
break;
}
}
if (!np)
return NULL;
path = p;
}
/* Step down the tree matching path components */
raw_spin_lock_irqsave(&devtree_lock, flags);
if (!np)
np = of_node_get(of_root);
np = __of_find_node_by_full_path(np, path);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_node_opts_by_path);
/**
* of_find_node_by_name - Find a node by its "name" property
* @from: The node to start searching from or NULL; the node
* you pass will not be searched, only the next one
* will. Typically, you pass what the previous call
* returned. of_node_put() will be called on @from.
* @name: The name string to match against
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_node_by_name(struct device_node *from,
const char *name)
{
struct device_node *np;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
for_each_of_allnodes_from(from, np)
if (of_node_name_eq(np, name) && of_node_get(np))
break;
of_node_put(from);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_node_by_name);
/**
* of_find_node_by_type - Find a node by its "device_type" property
* @from: The node to start searching from, or NULL to start searching
* the entire device tree. The node you pass will not be
* searched, only the next one will; typically, you pass
* what the previous call returned. of_node_put() will be
* called on from for you.
* @type: The type string to match against
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_node_by_type(struct device_node *from,
const char *type)
{
struct device_node *np;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
for_each_of_allnodes_from(from, np)
if (__of_node_is_type(np, type) && of_node_get(np))
break;
of_node_put(from);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_node_by_type);
/**
* of_find_compatible_node - Find a node based on type and one of the
* tokens in its "compatible" property
* @from: The node to start searching from or NULL, the node
* you pass will not be searched, only the next one
* will; typically, you pass what the previous call
* returned. of_node_put() will be called on it
* @type: The type string to match "device_type" or NULL to ignore
* @compatible: The string to match to one of the tokens in the device
* "compatible" list.
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_compatible_node(struct device_node *from,
const char *type, const char *compatible)
{
struct device_node *np;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
for_each_of_allnodes_from(from, np)
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
if (__of_device_is_compatible(np, compatible, type, NULL) &&
of_node_get(np))
break;
of_node_put(from);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_compatible_node);
/**
* of_find_node_with_property - Find a node which has a property with
* the given name.
* @from: The node to start searching from or NULL, the node
* you pass will not be searched, only the next one
* will; typically, you pass what the previous call
* returned. of_node_put() will be called on it
* @prop_name: The name of the property to look for.
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_node_with_property(struct device_node *from,
const char *prop_name)
{
struct device_node *np;
struct property *pp;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
for_each_of_allnodes_from(from, np) {
for (pp = np->properties; pp; pp = pp->next) {
if (of_prop_cmp(pp->name, prop_name) == 0) {
of_node_get(np);
goto out;
}
}
}
out:
of_node_put(from);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_node_with_property);
static
const struct of_device_id *__of_match_node(const struct of_device_id *matches,
const struct device_node *node)
{
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
const struct of_device_id *best_match = NULL;
int score, best_score = 0;
if (!matches)
return NULL;
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
for (; matches->name[0] || matches->type[0] || matches->compatible[0]; matches++) {
score = __of_device_is_compatible(node, matches->compatible,
matches->type, matches->name);
if (score > best_score) {
best_match = matches;
best_score = score;
}
}
of: reimplement the matching method for __of_match_node() In the current implementation of __of_match_node(), it will compare each given match entry against all the node's compatible strings with of_device_is_compatible(). To achieve multiple compatible strings per node with ordering from specific to generic, this requires given matches to be ordered from specific to generic. For most of the drivers this is not true and also an alphabetical ordering is more sane there. Therefore, we define a following priority order for the match, and then scan all the entries to find the best match. 1. specific compatible && type && name 2. specific compatible && type 3. specific compatible && name 4. specific compatible 5. general compatible && type && name 6. general compatible && type 7. general compatible && name 8. general compatible 9. type && name 10. type 11. name v5: Fix nested locking bug v4: Short-circuit failure cases instead of mucking with score, and remove extra __of_device_is_compatible() wrapper stub. Move scoring logic directly into __of_device_is_compatible() v3: Also need to bail out when there does have a compatible member in match entry, but it doesn't match with the device node's compatible. v2: Fix the bug such as we get the same score for the following two match entries with the empty node 'name2 { };' struct of_device_id matches[] = { {.name = "name2", }, {.name = "name2", .type = "type1", }, {} }; Signed-off-by: Kevin Hao <haokexin@gmail.com> [grant.likely: added v4 changes] Signed-off-by: Grant Likely <grant.likely@linaro.org> Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com> Tested-by: Stephen Chivers <schivers@csc.com> Tested-by: Sachin Kamat <sachin.kamat@linaro.org>
2014-02-19 08:15:45 +00:00
return best_match;
}
/**
* of_match_node - Tell if a device_node has a matching of_match structure
* @matches: array of of device match structures to search in
* @node: the of device structure to match against
*
* Low level utility function used by device matching.
*/
const struct of_device_id *of_match_node(const struct of_device_id *matches,
const struct device_node *node)
{
const struct of_device_id *match;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
match = __of_match_node(matches, node);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return match;
}
EXPORT_SYMBOL(of_match_node);
/**
* of_find_matching_node_and_match - Find a node based on an of_device_id
* match table.
* @from: The node to start searching from or NULL, the node
* you pass will not be searched, only the next one
* will; typically, you pass what the previous call
* returned. of_node_put() will be called on it
* @matches: array of of device match structures to search in
* @match: Updated to point at the matches entry which matched
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_matching_node_and_match(struct device_node *from,
const struct of_device_id *matches,
const struct of_device_id **match)
{
struct device_node *np;
const struct of_device_id *m;
unsigned long flags;
if (match)
*match = NULL;
raw_spin_lock_irqsave(&devtree_lock, flags);
for_each_of_allnodes_from(from, np) {
m = __of_match_node(matches, np);
if (m && of_node_get(np)) {
if (match)
*match = m;
break;
}
}
of_node_put(from);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_matching_node_and_match);
/**
* of_alias_from_compatible - Lookup appropriate alias for a device node
* depending on compatible
* @node: pointer to a device tree node
* @alias: Pointer to buffer that alias value will be copied into
* @len: Length of alias value
*
* Based on the value of the compatible property, this routine will attempt
* to choose an appropriate alias value for a particular device tree node.
* It does this by stripping the manufacturer prefix (as delimited by a ',')
* from the first entry in the compatible list property.
*
* Note: The matching on just the "product" side of the compatible is a relic
* from I2C and SPI. Please do not add any new user.
*
* Return: This routine returns 0 on success, <0 on failure.
*/
int of_alias_from_compatible(const struct device_node *node, char *alias, int len)
{
const char *compatible, *p;
int cplen;
compatible = of_get_property(node, "compatible", &cplen);
if (!compatible || strlen(compatible) > cplen)
return -ENODEV;
p = strchr(compatible, ',');
strscpy(alias, p ? p + 1 : compatible, len);
return 0;
}
EXPORT_SYMBOL_GPL(of_alias_from_compatible);
/**
* of_find_node_by_phandle - Find a node given a phandle
* @handle: phandle of the node to find
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done.
*/
struct device_node *of_find_node_by_phandle(phandle handle)
{
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
struct device_node *np = NULL;
unsigned long flags;
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
u32 handle_hash;
if (!handle)
return NULL;
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
handle_hash = of_phandle_cache_hash(handle);
raw_spin_lock_irqsave(&devtree_lock, flags);
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
if (phandle_cache[handle_hash] &&
handle == phandle_cache[handle_hash]->phandle)
np = phandle_cache[handle_hash];
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
if (!np) {
for_each_of_allnodes(np)
if (np->phandle == handle &&
!of_node_check_flag(np, OF_DETACHED)) {
of: Rework and simplify phandle cache to use a fixed size The phandle cache was added to speed up of_find_node_by_phandle() by avoiding walking the whole DT to find a matching phandle. The implementation has several shortcomings: - The cache is designed to work on a linear set of phandle values. This is true for dtc generated DTs, but not for other cases such as Power. - The cache isn't enabled until of_core_init() and a typical system may see hundreds of calls to of_find_node_by_phandle() before that point. - The cache is freed and re-allocated when the number of phandles changes. - It takes a raw spinlock around a memory allocation which breaks on RT. Change the implementation to a fixed size and use hash_32() as the cache index. This greatly simplifies the implementation. It avoids the need for any re-alloc of the cache and taking a reference on nodes in the cache. We only have a single source of removing cache entries which is of_detach_node(). Using hash_32() removes any assumption on phandle values improving the hit rate for non-linear phandle values. The effect on linear values using hash_32() is about a 10% collision. The chances of thrashing on colliding values seems to be low. To compare performance, I used a RK3399 board which is a pretty typical system. I found that just measuring boot time as done previously is noisy and may be impacted by other things. Also bringing up secondary cores causes some issues with measuring, so I booted with 'nr_cpus=1'. With no caching, calls to of_find_node_by_phandle() take about 20124 us for 1248 calls. There's an additional 288 calls before time keeping is up. Using the average time per hit/miss with the cache, we can calculate these calls to take 690 us (277 hit / 11 miss) with a 128 entry cache and 13319 us with no cache or an uninitialized cache. Comparing the 3 implementations the time spent in of_find_node_by_phandle() is: no cache: 20124 us (+ 13319 us) 128 entry cache: 5134 us (+ 690 us) current cache: 819 us (+ 13319 us) We could move the allocation of the cache earlier to improve the current cache, but that just further complicates the situation as it needs to be after slab is up, so we can't do it when unflattening (which uses memblock). Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Segher Boessenkool <segher@kernel.crashing.org> Cc: Frank Rowand <frowand.list@gmail.com> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Reviewed-by: Frank Rowand <frowand.list@gmail.com> Tested-by: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Rob Herring <robh@kernel.org>
2019-12-06 22:27:41 +00:00
phandle_cache[handle_hash] = np;
of: cache phandle nodes to reduce cost of of_find_node_by_phandle() Create a cache of the nodes that contain a phandle property. Use this cache to find the node for a given phandle value instead of scanning the devicetree to find the node. If the phandle value is not found in the cache, of_find_node_by_phandle() will fall back to the tree scan algorithm. The cache is initialized in of_core_init(). The cache is freed via a late_initcall_sync() if modules are not enabled. If the devicetree is created by the dtc compiler, with all phandle property values auto generated, then the size required by the cache could be 4 * (1 + number of phandles) bytes. This results in an O(1) node lookup cost for a given phandle value. Due to a concern that the phandle property values might not be consistent with what is generated by the dtc compiler, a mask has been added to the cache lookup algorithm. To maintain the O(1) node lookup cost, the size of the cache has been increased by rounding the number of entries up to the next power of two. The overhead of finding the devicetree node containing a given phandle value has been noted by several people in the recent past, in some cases with a patch to add a hashed index of devicetree nodes, based on the phandle value of the node. One concern with this approach is the extra space added to each node. This patch takes advantage of the phandle property values auto generated by the dtc compiler, which begin with one and monotonically increase by one, resulting in a range of 1..n for n phandle values. This implementation should also provide a good reduction of overhead for any range of phandle values that are mostly in a monotonic range. Performance measurements by Chintan Pandya <cpandya@codeaurora.org> of several implementations of patches that are similar to this one suggest an expected reduction of boot time by ~400ms for his test system. If the cache size was decreased to 64 entries, the boot time was reduced by ~340 ms. The measurements were on a 4.9.73 kernel for arch/arm64/boot/dts/qcom/sda670-mtp.dts, contains 2371 nodes and 814 phandle values. Reported-by: Chintan Pandya <cpandya@codeaurora.org> Signed-off-by: Frank Rowand <frank.rowand@sony.com> Signed-off-by: Rob Herring <robh@kernel.org>
2018-03-05 00:14:47 +00:00
break;
}
}
of_node_get(np);
raw_spin_unlock_irqrestore(&devtree_lock, flags);
return np;
}
EXPORT_SYMBOL(of_find_node_by_phandle);
void of_print_phandle_args(const char *msg, const struct of_phandle_args *args)
{
int i;
printk("%s %pOF", msg, args->np);
for (i = 0; i < args->args_count; i++) {
const char delim = i ? ',' : ':';
pr_cont("%c%08x", delim, args->args[i]);
}
pr_cont("\n");
}
int of_phandle_iterator_init(struct of_phandle_iterator *it,
const struct device_node *np,
const char *list_name,
const char *cells_name,
int cell_count)
{
const __be32 *list;
int size;
memset(it, 0, sizeof(*it));
of: restore old handling of cells_name=NULL in of_*_phandle_with_args() Before commit e42ee61017f5 ("of: Let of_for_each_phandle fallback to non-negative cell_count") the iterator functions calling of_for_each_phandle assumed a cell count of 0 if cells_name was NULL. This corner case was missed when implementing the fallback logic in e42ee61017f5 and resulted in an endless loop. Restore the old behaviour of of_count_phandle_with_args() and of_parse_phandle_with_args() and add a check to of_phandle_iterator_init() to prevent a similar failure as a safety precaution. of_parse_phandle_with_args_map() doesn't need a similar fix as cells_name isn't NULL there. Affected drivers are: - drivers/base/power/domain.c - drivers/base/power/domain.c - drivers/clk/ti/clk-dra7-atl.c - drivers/hwmon/ibmpowernv.c - drivers/i2c/muxes/i2c-demux-pinctrl.c - drivers/iommu/mtk_iommu.c - drivers/net/ethernet/freescale/fman/mac.c - drivers/opp/of.c - drivers/perf/arm_dsu_pmu.c - drivers/regulator/of_regulator.c - drivers/remoteproc/imx_rproc.c - drivers/soc/rockchip/pm_domains.c - sound/soc/fsl/imx-audmix.c - sound/soc/fsl/imx-audmix.c - sound/soc/meson/axg-card.c - sound/soc/samsung/tm2_wm5110.c - sound/soc/samsung/tm2_wm5110.c Thanks to Geert Uytterhoeven for reporting the issue, Peter Rosin for helping pinpoint the actual problem and the testers for confirming this fix. Fixes: e42ee61017f5 ("of: Let of_for_each_phandle fallback to non-negative cell_count") Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Signed-off-by: Rob Herring <robh@kernel.org>
2019-09-18 08:47:48 +00:00
/*
* one of cell_count or cells_name must be provided to determine the
* argument length.
*/
if (cell_count < 0 && !cells_name)
return -EINVAL;
list = of_get_property(np, list_name, &size);
if (!list)
return -ENOENT;
it->cells_name = cells_name;
it->cell_count = cell_count;
it->parent = np;
it->list_end = list + size / sizeof(*list);
it->phandle_end = list;
it->cur = list;
return 0;
}
EXPORT_SYMBOL_GPL(of_phandle_iterator_init);
int of_phandle_iterator_next(struct of_phandle_iterator *it)
{
uint32_t count = 0;
if (it->node) {
of_node_put(it->node);
it->node = NULL;
}
if (!it->cur || it->phandle_end >= it->list_end)
return -ENOENT;
it->cur = it->phandle_end;
/* If phandle is 0, then it is an empty entry with no arguments. */
it->phandle = be32_to_cpup(it->cur++);
if (it->phandle) {
/*
* Find the provider node and parse the #*-cells property to
* determine the argument length.
*/
it->node = of_find_node_by_phandle(it->phandle);
if (it->cells_name) {
if (!it->node) {
pr_err("%pOF: could not find phandle %d\n",
it->parent, it->phandle);
goto err;
}
if (of_property_read_u32(it->node, it->cells_name,
&count)) {
/*
* If both cell_count and cells_name is given,
* fall back to cell_count in absence
* of the cells_name property
*/
if (it->cell_count >= 0) {
count = it->cell_count;
} else {
pr_err("%pOF: could not get %s for %pOF\n",
it->parent,
it->cells_name,
it->node);
goto err;
}
}
} else {
count = it->cell_count;
}
/*
* Make sure that the arguments actually fit in the remaining
* property data length
*/
if (it->cur + count > it->list_end) {
if (it->cells_name)
pr_err("%pOF: %s = %d found %td\n",
it->parent, it->cells_name,
count, it->list_end - it->cur);
else
pr_err("%pOF: phandle %s needs %d, found %td\n",
it->parent, of_node_full_name(it->node),
count, it->list_end - it->cur);
goto err;
}
}
it->phandle_end = it->cur + count;
it->cur_count = count;
return 0;
err:
if (it->node) {
of_node_put(it->node);
it->node = NULL;
}
return -EINVAL;
}
EXPORT_SYMBOL_GPL(of_phandle_iterator_next);
int of_phandle_iterator_args(struct of_phandle_iterator *it,
uint32_t *args,
int size)
{
int i, count;
count = it->cur_count;
if (WARN_ON(size < count))
count = size;
for (i = 0; i < count; i++)
args[i] = be32_to_cpup(it->cur++);
return count;
}
int __of_parse_phandle_with_args(const struct device_node *np,
const char *list_name,
const char *cells_name,
int cell_count, int index,
struct of_phandle_args *out_args)
{
struct of_phandle_iterator it;
int rc, cur_index = 0;
if (index < 0)
return -EINVAL;
/* Loop over the phandles until all the requested entry is found */
of_for_each_phandle(&it, rc, np, list_name, cells_name, cell_count) {
/*
* All of the error cases bail out of the loop, so at
* this point, the parsing is successful. If the requested
* index matches, then fill the out_args structure and return,
* or return -ENOENT for an empty entry.
*/
rc = -ENOENT;
if (cur_index == index) {
if (!it.phandle)
goto err;
if (out_args) {
int c;
c = of_phandle_iterator_args(&it,
out_args->args,
MAX_PHANDLE_ARGS);
out_args->np = it.node;
out_args->args_count = c;
} else {
of_node_put(it.node);
}
/* Found it! return success */
return 0;
}
cur_index++;
}
/*
* Unlock node before returning result; will be one of:
* -ENOENT : index is for empty phandle
* -EINVAL : parsing error on data
*/
err:
of_node_put(it.node);
return rc;
}
EXPORT_SYMBOL(__of_parse_phandle_with_args);
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
/**
* of_parse_phandle_with_args_map() - Find a node pointed by phandle in a list and remap it
* @np: pointer to a device tree node containing a list
* @list_name: property name that contains a list
* @stem_name: stem of property names that specify phandles' arguments count
* @index: index of a phandle to parse out
* @out_args: optional pointer to output arguments structure (will be filled)
*
* This function is useful to parse lists of phandles and their arguments.
* Returns 0 on success and fills out_args, on error returns appropriate errno
* value. The difference between this function and of_parse_phandle_with_args()
* is that this API remaps a phandle if the node the phandle points to has
* a <@stem_name>-map property.
*
* Caller is responsible to call of_node_put() on the returned out_args->np
* pointer.
*
* Example::
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
*
* phandle1: node1 {
* #list-cells = <2>;
* };
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
*
* phandle2: node2 {
* #list-cells = <1>;
* };
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
*
* phandle3: node3 {
* #list-cells = <1>;
* list-map = <0 &phandle2 3>,
* <1 &phandle2 2>,
* <2 &phandle1 5 1>;
* list-map-mask = <0x3>;
* };
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
*
* node4 {
* list = <&phandle1 1 2 &phandle3 0>;
* };
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
*
* To get a device_node of the ``node2`` node you may call this:
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
* of_parse_phandle_with_args(node4, "list", "list", 1, &args);
*/
int of_parse_phandle_with_args_map(const struct device_node *np,
const char *list_name,
const char *stem_name,
int index, struct of_phandle_args *out_args)
{
char *cells_name, *map_name = NULL, *mask_name = NULL;
char *pass_name = NULL;
struct device_node *cur, *new = NULL;
const __be32 *map, *mask, *pass;
static const __be32 dummy_mask[] = { [0 ... MAX_PHANDLE_ARGS] = ~0 };
static const __be32 dummy_pass[] = { [0 ... MAX_PHANDLE_ARGS] = 0 };
__be32 initial_match_array[MAX_PHANDLE_ARGS];
const __be32 *match_array = initial_match_array;
int i, ret, map_len, match;
u32 list_size, new_size;
if (index < 0)
return -EINVAL;
cells_name = kasprintf(GFP_KERNEL, "#%s-cells", stem_name);
if (!cells_name)
return -ENOMEM;
ret = -ENOMEM;
map_name = kasprintf(GFP_KERNEL, "%s-map", stem_name);
if (!map_name)
goto free;
mask_name = kasprintf(GFP_KERNEL, "%s-map-mask", stem_name);
if (!mask_name)
goto free;
pass_name = kasprintf(GFP_KERNEL, "%s-map-pass-thru", stem_name);
if (!pass_name)
goto free;
ret = __of_parse_phandle_with_args(np, list_name, cells_name, -1, index,
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
out_args);
if (ret)
goto free;
/* Get the #<list>-cells property */
cur = out_args->np;
ret = of_property_read_u32(cur, cells_name, &list_size);
if (ret < 0)
goto put;
/* Precalculate the match array - this simplifies match loop */
for (i = 0; i < list_size; i++)
initial_match_array[i] = cpu_to_be32(out_args->args[i]);
ret = -EINVAL;
while (cur) {
/* Get the <list>-map property */
map = of_get_property(cur, map_name, &map_len);
if (!map) {
ret = 0;
goto free;
}
map_len /= sizeof(u32);
/* Get the <list>-map-mask property (optional) */
mask = of_get_property(cur, mask_name, NULL);
if (!mask)
mask = dummy_mask;
/* Iterate through <list>-map property */
match = 0;
while (map_len > (list_size + 1) && !match) {
/* Compare specifiers */
match = 1;
for (i = 0; i < list_size; i++, map_len--)
match &= !((match_array[i] ^ *map++) & mask[i]);
of_node_put(new);
new = of_find_node_by_phandle(be32_to_cpup(map));
map++;
map_len--;
/* Check if not found */
if (!new)
goto put;
if (!of_device_is_available(new))
match = 0;
ret = of_property_read_u32(new, cells_name, &new_size);
if (ret)
goto put;
/* Check for malformed properties */
if (WARN_ON(new_size > MAX_PHANDLE_ARGS))
goto put;
if (map_len < new_size)
goto put;
/* Move forward by new node's #<list>-cells amount */
map += new_size;
map_len -= new_size;
}
if (!match)
goto put;
/* Get the <list>-map-pass-thru property (optional) */
pass = of_get_property(cur, pass_name, NULL);
if (!pass)
pass = dummy_pass;
/*
* Successfully parsed a <list>-map translation; copy new
* specifier into the out_args structure, keeping the
* bits specified in <list>-map-pass-thru.
*/
match_array = map - new_size;
for (i = 0; i < new_size; i++) {
__be32 val = *(map - new_size + i);
if (i < list_size) {
val &= ~pass[i];
val |= cpu_to_be32(out_args->args[i]) & pass[i];
}
out_args->args[i] = be32_to_cpu(val);
}
out_args->args_count = list_size = new_size;
/* Iterate again with new provider */
out_args->np = new;
of_node_put(cur);
cur = new;
new = NULL;
of: Support parsing phandle argument lists through a nexus node Platforms like 96boards have a standardized connector/expansion slot that exposes signals like GPIOs to expansion boards in an SoC agnostic way. We'd like the DT overlays for the expansion boards to be written once without knowledge of the SoC on the other side of the connector. This avoids the unscalable combinatorial explosion of a different DT overlay for each expansion board and SoC pair. We need a way to describe the GPIOs routed through the connector in an SoC agnostic way. Let's introduce nexus property parsing into the OF core to do this. This is largely based on the interrupt nexus support we already have. This allows us to remap a phandle list in a consumer node (e.g. reset-gpios) through a connector in a generic way (e.g. via gpio-map). Do this in a generic routine so that we can remap any sort of variable length phandle list. Taking GPIOs as an example, the connector would be a GPIO nexus, supporting the remapping of a GPIO specifier space to multiple GPIO providers on the SoC. DT would look as shown below, where 'soc_gpio1' and 'soc_gpio2' are inside the SoC, 'connector' is an expansion port where boards can be plugged in, and 'expansion_device' is a device on the expansion board. soc { soc_gpio1: gpio-controller1 { #gpio-cells = <2>; }; soc_gpio2: gpio-controller2 { #gpio-cells = <2>; }; }; connector: connector { #gpio-cells = <2>; gpio-map = <0 0 &soc_gpio1 1 0>, <1 0 &soc_gpio2 4 0>, <2 0 &soc_gpio1 3 0>, <3 0 &soc_gpio2 2 0>; gpio-map-mask = <0xf 0x0>; gpio-map-pass-thru = <0x0 0x1> }; expansion_device { reset-gpios = <&connector 2 GPIO_ACTIVE_LOW>; }; The GPIO core would use of_parse_phandle_with_args_map() instead of of_parse_phandle_with_args() and arrive at the same type of result, a phandle and argument list. The difference is that the phandle and arguments will be remapped through the nexus node to the underlying SoC GPIO controller node. In the example above, we would remap 'reset-gpios' from <&connector 2 GPIO_ACTIVE_LOW> to <&soc_gpio1 3 GPIO_ACTIVE_LOW>. Cc: Pantelis Antoniou <pantelis.antoniou@konsulko.com> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: Mark Brown <broonie@kernel.org> Signed-off-by: Stephen Boyd <stephen.boyd@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org>
2018-01-31 02:36:16 +00:00
}
put:
of_node_put(cur);
of_node_put(new);
free:
kfree(mask_name);
kfree(map_name);
kfree(cells_name);
kfree(pass_name);
return ret;
}
EXPORT_SYMBOL(of_parse_phandle_with_args_map);
/**
* of_count_phandle_with_args() - Find the number of phandles references in a property
* @np: pointer to a device tree node containing a list
* @list_name: property name that contains a list
* @cells_name: property name that specifies phandles' arguments count
*
* Return: The number of phandle + argument tuples within a property. It
* is a typical pattern to encode a list of phandle and variable
* arguments into a single property. The number of arguments is encoded
* by a property in the phandle-target node. For example, a gpios
* property would contain a list of GPIO specifies consisting of a
* phandle and 1 or more arguments. The number of arguments are
* determined by the #gpio-cells property in the node pointed to by the
* phandle.
*/
int of_count_phandle_with_args(const struct device_node *np, const char *list_name,
const char *cells_name)
{
struct of_phandle_iterator it;
int rc, cur_index = 0;
of: restore old handling of cells_name=NULL in of_*_phandle_with_args() Before commit e42ee61017f5 ("of: Let of_for_each_phandle fallback to non-negative cell_count") the iterator functions calling of_for_each_phandle assumed a cell count of 0 if cells_name was NULL. This corner case was missed when implementing the fallback logic in e42ee61017f5 and resulted in an endless loop. Restore the old behaviour of of_count_phandle_with_args() and of_parse_phandle_with_args() and add a check to of_phandle_iterator_init() to prevent a similar failure as a safety precaution. of_parse_phandle_with_args_map() doesn't need a similar fix as cells_name isn't NULL there. Affected drivers are: - drivers/base/power/domain.c - drivers/base/power/domain.c - drivers/clk/ti/clk-dra7-atl.c - drivers/hwmon/ibmpowernv.c - drivers/i2c/muxes/i2c-demux-pinctrl.c - drivers/iommu/mtk_iommu.c - drivers/net/ethernet/freescale/fman/mac.c - drivers/opp/of.c - drivers/perf/arm_dsu_pmu.c - drivers/regulator/of_regulator.c - drivers/remoteproc/imx_rproc.c - drivers/soc/rockchip/pm_domains.c - sound/soc/fsl/imx-audmix.c - sound/soc/fsl/imx-audmix.c - sound/soc/meson/axg-card.c - sound/soc/samsung/tm2_wm5110.c - sound/soc/samsung/tm2_wm5110.c Thanks to Geert Uytterhoeven for reporting the issue, Peter Rosin for helping pinpoint the actual problem and the testers for confirming this fix. Fixes: e42ee61017f5 ("of: Let of_for_each_phandle fallback to non-negative cell_count") Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Signed-off-by: Rob Herring <robh@kernel.org>
2019-09-18 08:47:48 +00:00
/*
* If cells_name is NULL we assume a cell count of 0. This makes
* counting the phandles trivial as each 32bit word in the list is a
* phandle and no arguments are to consider. So we don't iterate through
* the list but just use the length to determine the phandle count.
*/
if (!cells_name) {
const __be32 *list;
int size;
list = of_get_property(np, list_name, &size);
if (!list)
return -ENOENT;
return size / sizeof(*list);
}
rc = of_phandle_iterator_init(&it, np, list_name, cells_name, -1);
if (rc)
return rc;
while ((rc = of_phandle_iterator_next(&it)) == 0)
cur_index += 1;
if (rc != -ENOENT)
return rc;
return cur_index;
}
EXPORT_SYMBOL(of_count_phandle_with_args);
static struct property *__of_remove_property_from_list(struct property **list, struct property *prop)
{
struct property **next;
for (next = list; *next; next = &(*next)->next) {
if (*next == prop) {
*next = prop->next;
prop->next = NULL;
return prop;
}
}
return NULL;
}
/**
* __of_add_property - Add a property to a node without lock operations
of: base: Fix some formatting issues and provide missing descriptions Fixes the following W=1 kernel build warning(s): drivers/of/base.c:315: warning: Function parameter or member 'cpun' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'prop_name' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'cpu' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'thread' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: expecting prototype for property holds the physical id of the(). Prototype was for __of_find_n_match_cpu_property() instead drivers/of/base.c:1139: warning: Function parameter or member 'match' not described in 'of_find_matching_node_and_match' drivers/of/base.c:1779: warning: Function parameter or member 'np' not described in '__of_add_property' drivers/of/base.c:1779: warning: Function parameter or member 'prop' not described in '__of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'np' not described in 'of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'prop' not described in 'of_add_property' drivers/of/base.c:1849: warning: Function parameter or member 'np' not described in 'of_remove_property' drivers/of/base.c:1849: warning: Function parameter or member 'prop' not described in 'of_remove_property' drivers/of/base.c:2137: warning: Function parameter or member 'dn' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'name' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'index' not described in 'of_console_check' Cc: Rob Herring <robh+dt@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: devicetree@vger.kernel.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20210318104036.3175910-5-lee.jones@linaro.org
2021-03-18 10:40:30 +00:00
* @np: Caller's Device Node
* @prop: Property to add
*/
int __of_add_property(struct device_node *np, struct property *prop)
{
int rc = 0;
unsigned long flags;
struct property **next;
raw_spin_lock_irqsave(&devtree_lock, flags);
__of_remove_property_from_list(&np->deadprops, prop);
prop->next = NULL;
next = &np->properties;
while (*next) {
if (strcmp(prop->name, (*next)->name) == 0) {
/* duplicate ! don't insert it */
rc = -EEXIST;
goto out_unlock;
}
next = &(*next)->next;
}
*next = prop;
out_unlock:
raw_spin_unlock_irqrestore(&devtree_lock, flags);
if (rc)
return rc;
__of_add_property_sysfs(np, prop);
return 0;
}
/**
* of_add_property - Add a property to a node
of: base: Fix some formatting issues and provide missing descriptions Fixes the following W=1 kernel build warning(s): drivers/of/base.c:315: warning: Function parameter or member 'cpun' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'prop_name' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'cpu' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'thread' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: expecting prototype for property holds the physical id of the(). Prototype was for __of_find_n_match_cpu_property() instead drivers/of/base.c:1139: warning: Function parameter or member 'match' not described in 'of_find_matching_node_and_match' drivers/of/base.c:1779: warning: Function parameter or member 'np' not described in '__of_add_property' drivers/of/base.c:1779: warning: Function parameter or member 'prop' not described in '__of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'np' not described in 'of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'prop' not described in 'of_add_property' drivers/of/base.c:1849: warning: Function parameter or member 'np' not described in 'of_remove_property' drivers/of/base.c:1849: warning: Function parameter or member 'prop' not described in 'of_remove_property' drivers/of/base.c:2137: warning: Function parameter or member 'dn' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'name' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'index' not described in 'of_console_check' Cc: Rob Herring <robh+dt@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: devicetree@vger.kernel.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20210318104036.3175910-5-lee.jones@linaro.org
2021-03-18 10:40:30 +00:00
* @np: Caller's Device Node
* @prop: Property to add
*/
int of_add_property(struct device_node *np, struct property *prop)
{
int rc;
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_lock(&of_mutex);
rc = __of_add_property(np, prop);
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_unlock(&of_mutex);
if (!rc)
of_property_notify(OF_RECONFIG_ADD_PROPERTY, np, prop, NULL);
return rc;
}
EXPORT_SYMBOL_GPL(of_add_property);
int __of_remove_property(struct device_node *np, struct property *prop)
{
unsigned long flags;
int rc = -ENODEV;
raw_spin_lock_irqsave(&devtree_lock, flags);
if (__of_remove_property_from_list(&np->properties, prop)) {
/* Found the property, add it to deadprops list */
prop->next = np->deadprops;
np->deadprops = prop;
rc = 0;
}
raw_spin_unlock_irqrestore(&devtree_lock, flags);
if (rc)
return rc;
__of_remove_property_sysfs(np, prop);
return 0;
}
/**
* of_remove_property - Remove a property from a node.
of: base: Fix some formatting issues and provide missing descriptions Fixes the following W=1 kernel build warning(s): drivers/of/base.c:315: warning: Function parameter or member 'cpun' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'prop_name' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'cpu' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'thread' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: expecting prototype for property holds the physical id of the(). Prototype was for __of_find_n_match_cpu_property() instead drivers/of/base.c:1139: warning: Function parameter or member 'match' not described in 'of_find_matching_node_and_match' drivers/of/base.c:1779: warning: Function parameter or member 'np' not described in '__of_add_property' drivers/of/base.c:1779: warning: Function parameter or member 'prop' not described in '__of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'np' not described in 'of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'prop' not described in 'of_add_property' drivers/of/base.c:1849: warning: Function parameter or member 'np' not described in 'of_remove_property' drivers/of/base.c:1849: warning: Function parameter or member 'prop' not described in 'of_remove_property' drivers/of/base.c:2137: warning: Function parameter or member 'dn' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'name' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'index' not described in 'of_console_check' Cc: Rob Herring <robh+dt@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: devicetree@vger.kernel.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20210318104036.3175910-5-lee.jones@linaro.org
2021-03-18 10:40:30 +00:00
* @np: Caller's Device Node
* @prop: Property to remove
*
* Note that we don't actually remove it, since we have given out
* who-knows-how-many pointers to the data using get-property.
* Instead we just move the property to the "dead properties"
* list, so it won't be found any more.
*/
int of_remove_property(struct device_node *np, struct property *prop)
{
int rc;
if (!prop)
return -ENODEV;
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_lock(&of_mutex);
rc = __of_remove_property(np, prop);
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_unlock(&of_mutex);
if (!rc)
of_property_notify(OF_RECONFIG_REMOVE_PROPERTY, np, prop, NULL);
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
return rc;
}
EXPORT_SYMBOL_GPL(of_remove_property);
int __of_update_property(struct device_node *np, struct property *newprop,
struct property **oldpropp)
{
struct property **next, *oldprop;
unsigned long flags;
raw_spin_lock_irqsave(&devtree_lock, flags);
__of_remove_property_from_list(&np->deadprops, newprop);
for (next = &np->properties; *next; next = &(*next)->next) {
if (of_prop_cmp((*next)->name, newprop->name) == 0)
break;
}
*oldpropp = oldprop = *next;
if (oldprop) {
/* replace the node */
newprop->next = oldprop->next;
*next = newprop;
oldprop->next = np->deadprops;
np->deadprops = oldprop;
} else {
/* new node */
newprop->next = NULL;
*next = newprop;
}
raw_spin_unlock_irqrestore(&devtree_lock, flags);
__of_update_property_sysfs(np, newprop, oldprop);
return 0;
}
/*
* of_update_property - Update a property in a node, if the property does
* not exist, add it.
*
* Note that we don't actually remove it, since we have given out
* who-knows-how-many pointers to the data using get-property.
* Instead we just move the property to the "dead properties" list,
* and add the new property to the property list
*/
int of_update_property(struct device_node *np, struct property *newprop)
{
struct property *oldprop;
int rc;
if (!newprop->name)
return -EINVAL;
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_lock(&of_mutex);
rc = __of_update_property(np, newprop, &oldprop);
of: Make devicetree sysfs update functions consistent. All of the DT modification functions are split into two parts, the first part manipulates the DT data structure, and the second part updates sysfs, but the code isn't very consistent about how the second half is called. They don't all enforce the same rules about when it is valid to update sysfs, and there isn't any clarity on locking. The transactional DT modification feature that is coming also needs access to these functions so that it can perform all the structure changes together, and then all the sysfs updates as a second stage instead of doing each one at a time. Fix up the second have by creating a separate __of_*_sysfs() function for each of the helpers. The new functions have consistent naming (ie. of_node_add() becomes __of_attach_node_sysfs()) and all of them now defer if of_init hasn't been called yet. Callers of the new functions must hold the of_mutex to ensure there are no race conditions with of_init(). The mutex ensures that there will only ever be one writer to the tree at any given time. There can still be any number of readers and the raw_spin_lock is still used to make sure access to the data structure is still consistent. Finally, put the function prototypes into of_private.h so they are accessible to the transaction code. Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com> [grant.likely: Changed suffix from _post to _sysfs to match existing code] [grant.likely: Reorganized to eliminate trivial wrappers] Signed-off-by: Grant Likely <grant.likely@linaro.org>
2014-07-23 23:05:06 +00:00
mutex_unlock(&of_mutex);
if (!rc)
of_property_notify(OF_RECONFIG_UPDATE_PROPERTY, np, newprop, oldprop);
return rc;
}
static void of_alias_add(struct alias_prop *ap, struct device_node *np,
int id, const char *stem, int stem_len)
{
ap->np = np;
ap->id = id;
strscpy(ap->stem, stem, stem_len + 1);
list_add_tail(&ap->link, &aliases_lookup);
pr_debug("adding DT alias:%s: stem=%s id=%i node=%pOF\n",
ap->alias, ap->stem, ap->id, np);
}
/**
* of_alias_scan - Scan all properties of the 'aliases' node
* @dt_alloc: An allocator that provides a virtual address to memory
* for storing the resulting tree
*
* The function scans all the properties of the 'aliases' node and populates
* the global lookup table with the properties. It returns the
* number of alias properties found, or an error code in case of failure.
*/
void of_alias_scan(void * (*dt_alloc)(u64 size, u64 align))
{
struct property *pp;
of_aliases = of_find_node_by_path("/aliases");
of_chosen = of_find_node_by_path("/chosen");
if (of_chosen == NULL)
of_chosen = of_find_node_by_path("/chosen@0");
if (of_chosen) {
/* linux,stdout-path and /aliases/stdout are for legacy compatibility */
const char *name = NULL;
if (of_property_read_string(of_chosen, "stdout-path", &name))
of_property_read_string(of_chosen, "linux,stdout-path",
&name);
if (IS_ENABLED(CONFIG_PPC) && !name)
of_property_read_string(of_aliases, "stdout", &name);
if (name)
of_stdout = of_find_node_opts_by_path(name, &of_stdout_options);
if (of_stdout)
of_stdout->fwnode.flags |= FWNODE_FLAG_BEST_EFFORT;
}
if (!of_aliases)
return;
for_each_property_of_node(of_aliases, pp) {
const char *start = pp->name;
const char *end = start + strlen(start);
struct device_node *np;
struct alias_prop *ap;
int id, len;
/* Skip those we do not want to proceed */
if (!strcmp(pp->name, "name") ||
!strcmp(pp->name, "phandle") ||
!strcmp(pp->name, "linux,phandle"))
continue;
np = of_find_node_by_path(pp->value);
if (!np)
continue;
/* walk the alias backwards to extract the id and work out
* the 'stem' string */
while (isdigit(*(end-1)) && end > start)
end--;
len = end - start;
if (kstrtoint(end, 10, &id) < 0)
continue;
/* Allocate an alias_prop with enough space for the stem */
ap = dt_alloc(sizeof(*ap) + len + 1, __alignof__(*ap));
if (!ap)
continue;
memset(ap, 0, sizeof(*ap) + len + 1);
ap->alias = start;
of_alias_add(ap, np, id, start, len);
}
}
/**
* of_alias_get_id - Get alias id for the given device_node
* @np: Pointer to the given device_node
* @stem: Alias stem of the given device_node
*
* The function travels the lookup table to get the alias id for the given
* device_node and alias stem.
*
* Return: The alias id if found.
*/
int of_alias_get_id(struct device_node *np, const char *stem)
{
struct alias_prop *app;
int id = -ENODEV;
mutex_lock(&of_mutex);
list_for_each_entry(app, &aliases_lookup, link) {
if (strcmp(app->stem, stem) != 0)
continue;
if (np == app->np) {
id = app->id;
break;
}
}
mutex_unlock(&of_mutex);
return id;
}
EXPORT_SYMBOL_GPL(of_alias_get_id);
/**
* of_alias_get_highest_id - Get highest alias id for the given stem
* @stem: Alias stem to be examined
*
* The function travels the lookup table to get the highest alias id for the
* given alias stem. It returns the alias id if found.
*/
int of_alias_get_highest_id(const char *stem)
{
struct alias_prop *app;
int id = -ENODEV;
mutex_lock(&of_mutex);
list_for_each_entry(app, &aliases_lookup, link) {
if (strcmp(app->stem, stem) != 0)
continue;
if (app->id > id)
id = app->id;
}
mutex_unlock(&of_mutex);
return id;
}
EXPORT_SYMBOL_GPL(of_alias_get_highest_id);
/**
* of_console_check() - Test and setup console for DT setup
of: base: Fix some formatting issues and provide missing descriptions Fixes the following W=1 kernel build warning(s): drivers/of/base.c:315: warning: Function parameter or member 'cpun' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'prop_name' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'cpu' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: Function parameter or member 'thread' not described in '__of_find_n_match_cpu_property' drivers/of/base.c:315: warning: expecting prototype for property holds the physical id of the(). Prototype was for __of_find_n_match_cpu_property() instead drivers/of/base.c:1139: warning: Function parameter or member 'match' not described in 'of_find_matching_node_and_match' drivers/of/base.c:1779: warning: Function parameter or member 'np' not described in '__of_add_property' drivers/of/base.c:1779: warning: Function parameter or member 'prop' not described in '__of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'np' not described in 'of_add_property' drivers/of/base.c:1800: warning: Function parameter or member 'prop' not described in 'of_add_property' drivers/of/base.c:1849: warning: Function parameter or member 'np' not described in 'of_remove_property' drivers/of/base.c:1849: warning: Function parameter or member 'prop' not described in 'of_remove_property' drivers/of/base.c:2137: warning: Function parameter or member 'dn' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'name' not described in 'of_console_check' drivers/of/base.c:2137: warning: Function parameter or member 'index' not described in 'of_console_check' Cc: Rob Herring <robh+dt@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: devicetree@vger.kernel.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Rob Herring <robh@kernel.org> Link: https://lore.kernel.org/r/20210318104036.3175910-5-lee.jones@linaro.org
2021-03-18 10:40:30 +00:00
* @dn: Pointer to device node
* @name: Name to use for preferred console without index. ex. "ttyS"
* @index: Index to use for preferred console.
*
* Check if the given device node matches the stdout-path property in the
* /chosen node. If it does then register it as the preferred console.
*
* Return: TRUE if console successfully setup. Otherwise return FALSE.
*/
bool of_console_check(struct device_node *dn, char *name, int index)
{
if (!dn || dn != of_stdout || console_set_on_cmdline)
return false;
/*
* XXX: cast `options' to char pointer to suppress complication
* warnings: printk, UART and console drivers expect char pointer.
*/
return !add_preferred_console(name, index, (char *)of_stdout_options);
}
EXPORT_SYMBOL_GPL(of_console_check);
/**
* of_find_next_cache_node - Find a node's subsidiary cache
* @np: node of type "cpu" or "cache"
*
* Return: A node pointer with refcount incremented, use
* of_node_put() on it when done. Caller should hold a reference
* to np.
*/
struct device_node *of_find_next_cache_node(const struct device_node *np)
{
struct device_node *child, *cache_node;
cache_node = of_parse_phandle(np, "l2-cache", 0);
if (!cache_node)
cache_node = of_parse_phandle(np, "next-level-cache", 0);
if (cache_node)
return cache_node;
/* OF on pmac has nodes instead of properties named "l2-cache"
* beneath CPU nodes.
*/
if (IS_ENABLED(CONFIG_PPC_PMAC) && of_node_is_type(np, "cpu"))
for_each_child_of_node(np, child)
if (of_node_is_type(child, "cache"))
return child;
return NULL;
}
/**
* of_find_last_cache_level - Find the level at which the last cache is
* present for the given logical cpu
*
* @cpu: cpu number(logical index) for which the last cache level is needed
*
* Return: The level at which the last cache is present. It is exactly
* same as the total number of cache levels for the given logical cpu.
*/
int of_find_last_cache_level(unsigned int cpu)
{
u32 cache_level = 0;
struct device_node *prev = NULL, *np = of_cpu_device_node_get(cpu);
while (np) {
of_node_put(prev);
prev = np;
np = of_find_next_cache_node(np);
}
of_property_read_u32(prev, "cache-level", &cache_level);
of_node_put(prev);
return cache_level;
}
/**
* of_map_id - Translate an ID through a downstream mapping.
* @np: root complex device node.
* @id: device ID to map.
* @map_name: property name of the map to use.
* @map_mask_name: optional property name of the mask to use.
* @target: optional pointer to a target device node.
* @id_out: optional pointer to receive the translated ID.
*
* Given a device ID, look up the appropriate implementation-defined
* platform ID and/or the target device which receives transactions on that
* ID, as per the "iommu-map" and "msi-map" bindings. Either of @target or
* @id_out may be NULL if only the other is required. If @target points to
* a non-NULL device node pointer, only entries targeting that node will be
* matched; if it points to a NULL value, it will receive the device node of
* the first matching target phandle, with a reference held.
*
* Return: 0 on success or a standard error code on failure.
*/
int of_map_id(struct device_node *np, u32 id,
const char *map_name, const char *map_mask_name,
struct device_node **target, u32 *id_out)
{
u32 map_mask, masked_id;
int map_len;
const __be32 *map = NULL;
if (!np || !map_name || (!target && !id_out))
return -EINVAL;
map = of_get_property(np, map_name, &map_len);
if (!map) {
if (target)
return -ENODEV;
/* Otherwise, no map implies no translation */
*id_out = id;
return 0;
}
if (!map_len || map_len % (4 * sizeof(*map))) {
pr_err("%pOF: Error: Bad %s length: %d\n", np,
map_name, map_len);
return -EINVAL;
}
/* The default is to select all bits. */
map_mask = 0xffffffff;
/*
* Can be overridden by "{iommu,msi}-map-mask" property.
* If of_property_read_u32() fails, the default is used.
*/
if (map_mask_name)
of_property_read_u32(np, map_mask_name, &map_mask);
masked_id = map_mask & id;
for ( ; map_len > 0; map_len -= 4 * sizeof(*map), map += 4) {
struct device_node *phandle_node;
u32 id_base = be32_to_cpup(map + 0);
u32 phandle = be32_to_cpup(map + 1);
u32 out_base = be32_to_cpup(map + 2);
u32 id_len = be32_to_cpup(map + 3);
if (id_base & ~map_mask) {
pr_err("%pOF: Invalid %s translation - %s-mask (0x%x) ignores id-base (0x%x)\n",
np, map_name, map_name,
map_mask, id_base);
return -EFAULT;
}
if (masked_id < id_base || masked_id >= id_base + id_len)
continue;
phandle_node = of_find_node_by_phandle(phandle);
if (!phandle_node)
return -ENODEV;
if (target) {
if (*target)
of_node_put(phandle_node);
else
*target = phandle_node;
if (*target != phandle_node)
continue;
}
if (id_out)
*id_out = masked_id - id_base + out_base;
pr_debug("%pOF: %s, using mask %08x, id-base: %08x, out-base: %08x, length: %08x, id: %08x -> %08x\n",
np, map_name, map_mask, id_base, out_base,
id_len, id, masked_id - id_base + out_base);
return 0;
}
pr_info("%pOF: no %s translation for id 0x%x on %pOF\n", np, map_name,
id, target && *target ? *target : NULL);
/* Bypasses translation */
if (id_out)
*id_out = id;
return 0;
}
EXPORT_SYMBOL_GPL(of_map_id);