The new field 'dma_range_map' in struct device is used to facilitate the
use of single or multiple offsets between mapping regions of cpu addrs and
dma addrs. It subsumes the role of "dev->dma_pfn_offset" which was only
capable of holding a single uniform offset and had no region bounds
checking.
The function of_dma_get_range() has been modified so that it takes a single
argument -- the device node -- and returns a map, NULL, or an error code.
The map is an array that holds the information regarding the DMA regions.
Each range entry contains the address offset, the cpu_start address, the
dma_start address, and the size of the region.
of_dma_configure() is the typical manner to set range offsets but there are
a number of ad hoc assignments to "dev->dma_pfn_offset" in the kernel
driver code. These cases now invoke the function
dma_direct_set_offset(dev, cpu_addr, dma_addr, size).
Signed-off-by: Jim Quinlan <james.quinlan@broadcom.com>
[hch: various interface cleanups]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Rob Herring <robh@kernel.org>
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>
In some cases such as PCI host controllers, we may have a "parent bus"
which is an OF leaf node, but still need to correctly parse ranges from
the point of view of that bus. For that, factor out variants of the
"#addr-cells" and "#size-cells" parsers which do not assume they have a
device node and thus immediately traverse upwards before reading the
relevant property.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
[robh: don't make of_bus_n_{addr,size}_cells() public]
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Reviewed-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Signed-off-by: Rob Herring <robh@kernel.org>
of_dma_get_range() is only used within the DT core code, so remove the
export and move the header declaration to the private header.
Cc: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Reviewed-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Rob Herring <robh@kernel.org>
Non-overlay dynamic devicetree node removal may leave the node in
the phandle cache. Subsequent calls to of_find_node_by_phandle()
will incorrectly find the stale entry. Remove the node from the
cache.
Add paranoia checks in of_find_node_by_phandle() as a second level
of defense (do not return cached node if detached, do not add node
to cache if detached).
Fixes: 0b3ce78e90 ("of: cache phandle nodes to reduce cost of of_find_node_by_phandle()")
Reported-by: Michael Bringmann <mwb@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org # v4.17+
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Only some old OpenFirmware implementations rely on default sizes. Any
FDT and modern implementation should have explicit properties. Make the
OF_ROOT_NODE_*_CELLS_DEFAULT defines private so we don't get any outside
users.
This also gets us one step closer to removing the asm/prom.h dependency on
Sparc.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Frank Rowand <frowand.list@gmail.com>
Cc: sparclinux@vger.kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
A comment in the review of the patch adding the phandle cache said that
the cache would have to be updated when modules are applied and removed.
This patch implements the cache updates.
Fixes: 0b3ce78e90 ("of: cache phandle nodes to reduce cost of of_find_node_by_phandle()")
Reported-by: Alan Tull <atull@kernel.org>
Suggested-by: Alan Tull <atull@kernel.org>
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
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>
Struct device_node full_name no longer includes the full path name
when the devicetree is created from a flattened device tree (FDT).
The overlay node creation code was not modified to reflect this
change. Fix the node full_name generated by overlay code to contain
only the basename.
Unittests call an overlay internal function to create new nodes.
Fix up these calls to provide basename only instead of the full
path.
Fixes: a7e4cfb0a7 ("of/fdt: only store the device node basename
in full_name")
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
The "%pOF" printf format was recently added to print the
full name of a device tree node, with the intent of changing
the node full_name field to contain only the node name instead
of the full path of the node.
dup_and_fixup_symbol_prop() duplicates a property from the
"/__symbols__" node of an overlay device tree. The value
of each duplicated property must be fixed up to include
the full path of a node in the live device tree. The
current code uses the node's full_name for that purpose.
Update the code to use the "%pOF" printf format to
determine the node's full path.
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
The process of applying an overlay consists of:
- unflatten an overlay FDT (flattened device tree) into an
EDT (expanded device tree)
- fixup the phandle values in the overlay EDT to fit in a
range above the phandle values in the live device tree
- create the overlay changeset to reflect the contents of
the overlay EDT
- apply the overlay changeset, to modify the live device tree,
potentially changing the maximum phandle value in the live
device tree
There is currently no protection against two overlay applies
concurrently determining what range of phandle values are in use
in the live device tree, and subsequently changing that range.
Add a mutex to prevent multiple overlay applies from occurring
simultaneously.
Move of_resolve_phandles() into of_overlay_apply() so that it does not
have to be duplicated by each caller of of_overlay_apply().
The test in of_resolve_phandles() that the overlay tree is detached is
temporarily disabled so that old style overlay unittests do not fail.
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
When an attempt to apply an overlay changeset fails, an effort
is made to revert any partial application of the changeset.
When an attempt to remove an overlay changeset fails, an effort
is made to re-apply any partial reversion of the changeset.
The existing code does not check for failure to recover a failed
overlay changeset application or overlay changeset revert.
Add the missing checks and flag the devicetree as corrupt if the
state of the devicetree can not be determined.
Improve and expand the returned errors to more fully reflect the
result of the effort to undo the partial effects of a failed attempt
to apply or remove an overlay changeset.
If the device tree might be corrupt, do not allow further attempts
to apply or remove an overlay changeset.
When creating an overlay changeset from an overlay device tree,
add some additional warnings if the state of the overlay device
tree is not as expected.
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Having device_nodes be kobjects is only needed if sysfs or OF_DYNAMIC is
enabled. Otherwise, having a kobject in struct device_node is
unnecessary bloat in minimal kernel configurations.
Likewise, bin_attribute is only needed in struct property when sysfs is
enabled, so we can make it configurable too.
Tested-by: Nicolas Pitre <nico@linaro.org>
Reviewed-by: Frank Rowand <frowand.list@gmail.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Rob Herring <robh@kernel.org>
The only user of kobj_to_device_node() is in dynamic.c, so move it
there. This avoids having to make it conditional once kobject is
configurable.
Tested-by: Nicolas Pitre <nico@linaro.org>
Reviewed-by: Frank Rowand <frowand.list@gmail.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Rob Herring <robh@kernel.org>
find_node_by_full_name() does the same thing as of_find_node_by_path(),
but takes no locks and doesn't work on aliases. Refactor
of_find_node_opts_by_path() into __of_find_node_by_full_path() and
replace find_node_by_full_name() with it.
Signed-off-by: Rob Herring <robh@kernel.org>
__of_attach_node() is not used outside of drivers/of/dynamic.c. Make
it static and remove it from drivers/of/of_private.h.
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
We get a link error when the new tests are used by overlays
are not:
drivers/of/built-in.o: In function `unflatten_device_tree':
(.init.text+0x967): undefined reference to `unittest_unflatten_overlay_base'
This makes the #ifdef check match the symbols that lead to building
the unittest_unflatten_overlay_base function.
Fixes: 81d0848fc8 ("of: Add unit tests for applying overlays")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Frank Rowand <frank.rowand@sony.com>
Tested-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Existing overlay unit tests examine individual pieces of the overlay
code. The new tests target the entire process of applying an overlay.
Signed-off-by: Frank Rowand <frank.rowand@sony.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Fix a memory leak resulting from memory allocation in safe_name().
This patch fixes all call sites of safe_name().
Mathieu Malaterre reported the memory leak on boot:
On my PowerMac device-tree would generate a duplicate name:
[ 0.023043] device-tree: Duplicate name in PowerPC,G4@0, renamed to "l2-cache#1"
in this case a newly allocated name is generated by `safe_name`. However
in this case it is never deallocated.
The bug was found using kmemleak reported as:
unreferenced object 0xdf532e60 (size 32):
comm "swapper", pid 1, jiffies 4294892300 (age 1993.532s)
hex dump (first 32 bytes):
6c 32 2d 63 61 63 68 65 23 31 00 dd e4 dd 1e c2 l2-cache#1......
ec d4 ba ce 04 ec cc de 8e 85 e9 ca c4 ec cc 9e ................
backtrace:
[<c02d3350>] kvasprintf+0x64/0xc8
[<c02d3400>] kasprintf+0x4c/0x5c
[<c0453814>] safe_name.isra.1+0x80/0xc4
[<c04545d8>] __of_attach_node_sysfs+0x6c/0x11c
[<c075f21c>] of_core_init+0x8c/0xf8
[<c0729594>] kernel_init_freeable+0xd4/0x208
[<c00047e8>] kernel_init+0x24/0x11c
[<c00158ec>] ret_from_kernel_thread+0x5c/0x64
Link: https://bugzilla.kernel.org/show_bug.cgi?id=120331
Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Reported-by: mathieu.malaterre@gmail.com
Tested-by: Mathieu Malaterre <mathieu.malaterre@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
The PowerNV PCI hotplug driver is going to use the OF changeset
to manage the changed device sub-tree. This exports those OF
changeset functions for that.
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Acked-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Tested-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Add a node argument to __of_node_alloc() and rename it to
__of_node_dup() so that it can also be used to duplicate a node with
its properties. This is important for the overlay code so that it can
create new nodes without using separate changeset items for every single
property.
At the same time rework the overlay code to use the new function and
drop the extra changeset items.
Signed-off-by: Grant Likely <grant.likely@linaro.org>
The overlay code needs to construct a new full_name from the parent name
and the node name, but the current method has to allocate and then free
an temporary string which is wasteful. Fix this problem by using vargs
to pass in a format and arguments into __of_node_alloc().
At the same time remove the allocflags argument to __of_node_alloc().
The only users all use GFP_KERNEL, so there is no need to provide it as
an option. If there is ever a need later it can be added back.
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Introducing DT transactional support.
A DT transaction is a method which allows one to apply changes
in the live tree, in such a way that either the full set of changes
take effect, or the state of the tree can be rolled-back to the
state it was before it was attempted. An applied transaction
can be rolled-back at any time.
Documentation is in
Documentation/devicetree/changesets.txt
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
[glikely: Removed device notifiers and reworked to be more consistent]
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Currently, devicetree reconfig notifiers get emitted before the change
is applied to the tree, but that behaviour is problematic if the
receiver wants the determine the new state of the tree. The current
users don't care, but the changeset code to follow will be making
multiple changes at once. Reorder notifiers to get emitted after the
change has been applied to the tree so that callbacks see the new tree
state.
At the same time, fixup the existing callbacks to expect the new order.
There are a few callbacks that compare the old and new values of a
changed property. Put both property pointers into the of_prop_reconfig
structure.
The current notifiers also allow the notifier callback to fail and
cancel the change to the tree, but that feature isn't actually used.
It really isn't valid to ignore a tree modification provided by firmware
anyway, so remove the ability to cancel a change to the tree.
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Cc: Nathan Fontenot <nfont@austin.ibm.com>
PowerPC does an odd thing with dynamic nodes. It uses a notifier to
catch new node additions and set some of the values like name and type.
This makes no sense since that same code can be put directly into
of_attach_node(). Besides, all dynamic node users need this, not just
powerpc. Fix this problem by moving the logic out of arch/powerpc and
into drivers/of/dynamic.c.
It is also important to remove this notifier because we want to move the
firing of notifiers from before the tree is modified to after so that
the receiver gets a consistent view of the tree, but that is
incompatible with notifiers that modify the node.
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Cc: Nathan Fontenot <nfont@austin.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
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>
The DT overlay code will need to manipulate nodes and properties while
already holding the devicetree lock, or on nodes that are not yet
attached to the tree, but the current helper functions don't allow that.
Extract the core behaviour from the accessors and create the following
unlocked variants.
The unlocked variants require either the lock to already be held or for
the nodes to be detached from the tree. Changes to live nodes will not
get updated in sysfs, so the caller must arrange for housekeeping to
take place after dropping the lock.
The new functions are: __of_add_property(), __of_remove_property(),
__of_update_property(), __of_attach_node() and __of_detach_node().
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
[Remove unnecessary diff hunks and rewrite commit text]
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Introduce helper functions for working with the live DT tree,
all of them related to dynamically adding/removing nodes and
properties.
__of_prop_dup() copies a property dynamically
__of_node_alloc() creates an empty node
Bug fix about prop->len == 0 by Ionut Nicu <ioan.nicu.ext@nsn.com>
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
[glikely: Added unittest for of_copy_property and dropped fine-grained allocations]
[glikely: removed name, type and phandle arguments from __of_node_alloc]
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Split the dynamic device tree code into a separate file to make it
really clear what features CONFIF_OF_DYNAMIC add to the kernel. Without
CONFIG_OF_DYNAMIC only properties can be changed, and notifiers do not
get sent. Enabling it turns on reference counting, notifiers and the
ability to add and remove nodes.
v2: Moved of_node_release() into dynamic.c
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
Cc: Rob Herring <robh+dt@kernel.org>
We're overloading usage of of_aliases_mutex for sysfs changes,
so rename to something that is more generic.
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
Signed-off-by: Grant Likely <grant.likely@linaro.org>
In some situations, userspace may want to resolve a
device by function and logical number (ie, "serial0")
rather than by the base address or full device path. Being
able to resolve a device by alias frees userspace from the
burden of otherwise having to maintain a mapping between
device addresses and their logical assignments on each
platform when multiple instances of the same hardware block
are present in the system.
Although the uevent device attribute contains devicetree
compatible information and the full device path, the uevent
does not list the alises that may have been defined for the
device.
Signed-off-by: Stepan Moskovchenko <stepanm@codeaurora.org>
[grant.likely: Removed OF_ALIAS_N field; I don't think it's needed]
[grant.likely: Added #ifndef _LINUX_OF_PRIVATE_H wrapper]
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
