Add a function for collecting the amount of memory used by driver model,
including devices, uclasses and attached data and tags.
This information can provide insights into how to reduce the memory
required by driver model. Future work may look at execution speed also.
Signed-off-by: Simon Glass <sjg@chromium.org>
Introduce DM_FLAG_PROBE_AFTER_BIND flag, which can be set by driver or
uclass in .bind(), to indicate such driver instance should be probe()d
once binding of all devices is complete.
This is useful in case the driver determines that hardware initialization
is mandatory on boot, and such initialization happens only in probe().
This also solves the inability to call device_probe() from .bind().
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Cc: Sean Anderson <seanga2@gmail.com>
Cc: Simon Glass <sjg@chromium.org>
Cc: Steven Lawrance <steven.lawrance@softathome.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Tested-by: Patrice Chotard <patrice.chotard@foss.st.com>
When every member of a linker list is aligned by the compiler, we can no
longer rely on the sizeof of the struct to determine the number of
entries.
For example, if the struct size is 0x90 but every entry is aligned to 0xa0
by the compiler, the linker list entries takes more space in memory and
the calculation of the number of entries is incorrect. For example, we may
see 0x12 entries when there are only 0x11.
This is a real problem. There may be a general solution, although I cannot
currently think of one. So far it only bites with OF_PLATDATA_RT which
creates a pointer to each entry of the 'struct udevice' linker_list. This
does not happen without that option, so it only affects SPL.
Work around it by manually calculating the aligned size of struct udevice,
then using that for the n_ent calculation.
Note: the alignment fix to linker list was here:
0b2fa98aa5 linker_lists: Fix alignment issue
Signed-off-by: Simon Glass <sjg@chromium.org>
With dm-tag feature, any U-Boot subsystem is allowed to associate
arbitrary number of data with a particular udevice. This can been
see as expanding "struct udevice" without modifying the definition.
As a first user, UEFI subsystem makes use of tags to associate
an efi_disk object with a block device.
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Reviewed-by: Simon Glass <sjg@chromium.org>
Instead of a special function, send an event after driver model is inited
and adjust the boards which use this function.
Signed-off-by: Simon Glass <sjg@chromium.org>
- Disable bi_boot_params for STM32 MCU and MPU boards
- Update stm32-usbphyc node management
- Convert CONFIG_STM32_FLASH to Kconfig for STM32 MCU boards
- Convert some USB config flags to Kconfig for various boards
- Convert CONFIG_BOOTCOMMAND flag to Kconfig for STM32 F429 board
- Remove specific CONFIG_STV0991 flags
- Remove unused CONFIG_USER_LOWLEVEL_INIT flag
- Add ofdata_to_platdata() callback for stm32_spi driver
- Update for stm32f7_i2c driver
- Remove gpio_hog_probe_all() from STM32 MP1 board
- Fix bind command
-----BEGIN PGP SIGNATURE-----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=S4kl
-----END PGP SIGNATURE-----
Merge tag 'u-boot-stm32-20211012' of https://source.denx.de/u-boot/custodians/u-boot-stm
- Disable ATAGS for STM32 MCU and MPU boards
- Disable bi_boot_params for STM32 MCU and MPU boards
- Update stm32-usbphyc node management
- Convert CONFIG_STM32_FLASH to Kconfig for STM32 MCU boards
- Convert some USB config flags to Kconfig for various boards
- Convert CONFIG_BOOTCOMMAND flag to Kconfig for STM32 F429 board
- Remove specific CONFIG_STV0991 flags
- Remove unused CONFIG_USER_LOWLEVEL_INIT flag
- Add ofdata_to_platdata() callback for stm32_spi driver
- Update for stm32f7_i2c driver
- Remove gpio_hog_probe_all() from STM32 MP1 board
- Fix bind command
Signed-off-by: Tom Rini <trini@konsulko.com>
Fix a regression brings by commit 84f8e36f03 ("cmd: bind: allow to
bind driver with driver data")
As example, the following bind command doesn't work:
bind /soc/usb-otg@49000000 usb_ether
As usb_ether driver has no compatible string, it can't be find by
lists_bind_fdt(). In bind_by_node_path(), which called lists_bind_fdt(),
the driver entry is known, pass it to lists_bind_fdt() to force the driver
entry selection.
For this, add a new parameter struct *driver to lists_bind_fdt().
Fix also all lists_bind_fdt() callers.
Fixes: 84f8e36f03 ("cmd: bind: allow to bind driver with driver data")
Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reported-by: Herbert Poetzl <herbert@13thfloor.at>
Cc: Marek Vasut <marex@denx.de>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Each _device_ belonging to a given uclass of course has its own ->ops,
of a type determined by and known to the uclass.
However, no instance of a uclass_driver seems to populate ->ops, and
the only reference to it in code is this relocation.
Moreover, it's not really clear what could sensibly be assigned; it
would have to be some "struct uclass_ops *" providing a set of methods
for the core to call on that particular uclass, but should the need
for that ever arise, it would be better to have a member of that
particular type instead of void*.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Reviewed-by: Simon Glass <sjg@chromium.org>
Define LOG_CATEGORY for all uclass to allow filtering with
log command.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
If there are no nodes or if all nodes are disabled, this function would
return err without setting it first. Fix this by initializing err to
zero.
Fixes: 94f7afdf7e ("dm: core: Ignore disabled devices when binding")
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Make use of the new priv/plat data region if enabled. This is implemented
as a simple offset from the position set up by dtoc to the new position.
So long as all access goes through dm_priv_to_rw() this is safe.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Simon Glass <sjg@chromium.org>
At present the device priv/data data allocated by dtoc is stored in the
data section along with other variables. On some platforms it is better
to allocate space for it separately, e.g. if SPL is running from read-only
memory.
Create a new space with the same size as that allocated by dtoc, ready for
use.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Simon Glass <sjg@chromium.org>
At present when driver model needs to change a device it simply updates
the struct udevice structure. But with of-platdata-inst most of the fields
are not modified at runtime. In fact, typically only the flags need to
change.
For systems running SPL from read-only memory it is convenient to separate
out the runtime information, so that the devices don't need to be copied
before being used.
Create a new udevice_rt table, similar to the existing driver_rt. For now
it just holds the flags, although they are not used in this patch.
Add a new Kconfig for the driver_rt data, since this is not needed when
of-platdata-inst is used.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Simon Glass <sjg@chromium.org>
With this we don't need to scan and bind drivers, not even the root
device. We just need to locate the root device that was set up at build
time, then set our root in global_data to point to it.
Update the code to handle this case.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Simon Glass <sjg@chromium.org>
When OF_PLATDATA_INST is enabled we don't need to create the uclass list.
Instead we just need to point to the existing list. Update the code
accordingly.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Simon Glass <sjg@chromium.org>
Add another flag to the DM core which could be assigned to drivers and
which makes those drivers call their remove callbacks last, just before
booting OS and after all the other drivers finished with their remove
callbacks. This is necessary for things like clock drivers, where the
other drivers might depend on the clock driver in their remove callbacks.
Prime example is the mmc subsystem, which can reconfigure a card from HS
mode to slower modes in the remove callback and for that it needs to
reconfigure the controller clock.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com>
Signed-off-by: Simon Glass <sjg@chromium.org>
Move this out of the common header and include it only where needed. In
a number of cases this requires adding "struct udevice;" to avoid adding
another large header or in other cases replacing / adding missing header
files that had been pulled in, very indirectly. Finally, we have a few
cases where we did not need to include <asm/global_data.h> at all, so
remove that include.
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Tom Rini <trini@konsulko.com>
The __of_translate_address routine translates an address from the
device tree into a CPU physical address. A note in the description of
the routine explains that the crossing of any level with
since inherited from IBM. This does not happen for Texas Instruments, or
at least for the beaglebone device tree. Without this patch, in fact,
the translation into physical addresses of the registers contained in the
am33xx-clocks.dtsi nodes would not be possible. They all have a parent
with #size-cells = <0>.
The CONFIG_OF_TRANSLATE_ZERO_SIZE_CELLS symbol makes translation
possible even in the case of crossing levels with #size-cells = <0>.
The patch acts conservatively on address translation, except for
removing a check within the of_translate_one function in the
drivers/core/of_addr.c file:
+
ranges = of_get_property(parent, rprop, &rlen);
- if (ranges == NULL && !of_empty_ranges_quirk(parent)) {
- debug("no ranges; cannot translate\n");
- return 1;
- }
if (ranges == NULL || rlen == 0) {
offset = of_read_number(addr, na);
memset(addr, 0, pna * 4);
debug("empty ranges; 1:1 translation\n");
There are two reasons:
1 The function of_empty_ranges_quirk always returns false, invalidating
the following if statement in case of null ranges. Therefore one of
the two checks is useless.
2 The implementation of the of_translate_one function found in the
common/fdt_support.c file has removed this check while keeping the one
about the 1:1 translation.
The patch adds a test and modifies a check for the correctness of an
address in the case of enabling translation also for zero size cells.
The added test checks translations of addresses generated by nodes of
a device tree similar to those you can find in the files am33xx.dtsi
and am33xx-clocks.dtsi for which the patch was created.
The patch was also tested on a beaglebone black board. The addresses
generated for the registers of the loaded drivers are those specified
by the AM335x reference manual.
Signed-off-by: Dario Binacchi <dariobin@libero.it>
Tested-by: Dario Binacchi <dariobin@libero.it>
Reviewed-by: Simon Glass <sjg@chromium.org>
This has not been needed since parent information was added and we started
using indicies for references to other drivers instead of pointers. It was
kept around in the expectation that it might be needed later.
However with the latest updates, it doesn't seem likely that we'll need
this in the foreseeable future.
Drop dm_populate_phandle_data() from dtoc and driver model.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present the uclass list head is in global_data. This is convenient
but with the new of-platdata we need the list head to be declared by
the generated code.
Change this over to be a pointer. Provide a 'static' version in
global_data to retain the current behaviour.
Signed-off-by: Simon Glass <sjg@chromium.org>
Move the code related to scanning for devices to bind, into a new
function. This will make it easier to skip this step with the new
of-platdata improvements.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present ofnode is present in the device even if it is never used. With
of-platdata this field is not used, so can be removed. In preparation for
this, change the access to go through inline functions.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present the name 'uclass_driver' is used for the uclass linker list.
This does not follow the convention of using the struct name. Fix it.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present each device has two sequence numbers, with 'req_seq' being
set up at bind time and 'seq' at probe time. The idea is that devices
can 'request' a sequence number and then the conflicts are resolved when
the device is probed.
This makes things complicated in a few cases, since we don't really know
what the sequence number will end up being. We want to honour the
bind-time requests if at all possible, but in fact the only source of
these at present is the devicetree aliases. Since we have the devicetree
available at bind time, we may as well just use it, in the hope that the
required processing will turn out to be useful later (i.e. the device
actually gets used).
Add a new 'sqq' member, the bind-time sequence number. It operates in
parallel to the old values for now. All devices get a valid sqq value,
i.e. it is never -1.
Drop an #ifdef while we are here.
Signed-off-by: Simon Glass <sjg@chromium.org>
This name is far too long. Rename it to remove the 'data' bits. This makes
it consistent with the platdata->plat rename.
Signed-off-by: Simon Glass <sjg@chromium.org>
We use 'priv' for private data but often use 'platdata' for platform data.
We can't really use 'pdata' since that is ambiguous (it could mean private
or platform data).
Rename some of the latter variables to end with 'plat' for consistency.
Signed-off-by: Simon Glass <sjg@chromium.org>
This doesn't need to be passed the devicetree anymore. Drop it.
Also rename the function to drop the _fdt suffix.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present there are two copies of this code. With ofnode we can combine
them to reduce duplication. Update the dm_scan_fdt_node() function and
adjust its callers.
Signed-off-by: Simon Glass <sjg@chromium.org>
This is needed in at least one place. Avoid the conditional code in root.c
by adding this inline function.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present we update the driver_info struct with a pointer to the device
that it created (i.e. caused to be bound). This works fine when U-Boot SPL
is stored in read-write memory. But on some platforms, such as Intel
Apollo Lake, it is not possible to update the data memory.
In any case, it is bad form to put this information in a structure that is
in the data region, since it expands the size of the binary.
Create a new driver_rt structure which holds runtime information about
drivers. Update the code to store the device pointer in this instead.
Also update the test check that this works.
Signed-off-by: Simon Glass <sjg@chromium.org>
At present this option results in a number of #ifdefs due to the presence
or absence of the global_data of_root member.
Add a few macros to global_data.h to work around this. Update the code
accordingly.
Signed-off-by: Simon Glass <sjg@chromium.org>
This always has a fixed ACPI name so add it as a driver function.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Wolfgang Wallner <wolfgang.wallner@br-automation.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Currently when creating an U_BOOT_DEVICE entry a struct driver_info
is declared, which contains the data needed to instantiate the device.
However, the actual device is created at runtime and there is no proper
way to get the device based on its struct driver_info.
This patch extends struct driver_info adding a pointer to udevice which
is populated during the bind process, allowing to generate a set of
functions to get the device based on its struct driver_info.
Signed-off-by: Walter Lozano <walter.lozano@collabora.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
In order to prepare for a new support of phandle when OF_PLATDATA is used
drop the const for struct driver_info as this struct will need to be
updated on runtime.
Signed-off-by: Walter Lozano <walter.lozano@collabora.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Use the new function dm_scan_fdt_ofnode_path() to scan all the nodes
which aren't devices themselves but may contain some:
- "/chosen"
- "/clocks"
- "/firmware"
The patch removes the strcmp call in recursive function dm_scan_fdt_live()
and also corrects a conflict with the 2 applied patches in
the commit 1712ca2192 ("dm: core: Scan /firmware node by default")
and in the commit 747558d014 ("dm: fdt: scan for devices under
/firmware too"): the subnodes of "/firmware" (optee for example)
are bound 2 times.
For example the dm tree command result on STM32MP1 is:
STM32MP> dm tree
Class Index Probed Driver Name
-----------------------------------------------------------
root 0 [ + ] root_driver root_driver
firmware 0 [ ] psci |-- psci
sysreset 0 [ ] psci-sysreset | `-- psci-sysreset
simple_bus 0 [ + ] generic_simple_bus |-- soc
...
tee 0 [ + ] optee |-- optee
...
tee 1 [ ] optee `-- optee
Signed-off-by: Patrick Delaunay <patrick.delaunay@st.com>
Tested-by: Michal Simek <michal.simek@xilinx.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Some of the DM functions depend on OF_CONTROL, which is incorrect.
DM and DT are orthogonal. Add macro guards around such functions to
avoid compiling them in when DM is enabled, while OF_CONTROL is not.
Signed-off-by: Marek Vasut <marek.vasut+renesas@gmail.com>
Cc: Simon Glass <sjg@chromium.org>
Testing has shown that the current DM implementation of a platform /
board specific translation offset, as its needed for the SPL on MVEBU
platforms is buggy. The translation offset is confingured too late,
after the driver bind functions are run. This may result in incorrect
address translations. With the current implementation its not possible
to configure the offset earlier, as the DM code has not run at all.
This patch now removed the set_/get_translation_offset() calls and
moves the translation offset into the GD variable translation_offset.
This variable will get used when CONFIG_TRANSLATION_OFFSET is enabled.
This option is enabled only for MVEBU on ARM32 platforms, where its
currenty needed and configured in the SPL.
Signed-off-by: Stefan Roese <sr@denx.de>
Cc: Pierre Bourdon <delroth@gmail.com>
Cc: Baruch Siach <baruch@tkos.co.il>
Cc: Simon Glass <sjg@chromium.org>
Cc: Heiko Schocher <hs@denx.de>
Cc: Tom Rini <trini@konsulko.com>
Tested-by: Pierre Bourdon <delroth@gmail.com>
Tested-by: Baruch Siach <baruch@tkos.co.il>
To reset the DM after a new dtb is loaded, we need to call dm_uninit()
and then dm_init(). This fails however because gd->dm_root is not nullified
by dm_uninit().
Fixing it by setting gd->dm_root in dm_uninit().
Signed-off-by: Jean-Jacques Hiblot <jjhiblot@ti.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Heiko Schocher <hs@denx.de>
Commit f2006808f0: ("dm: core: parse chosen node") added a logic
to parse the chosen node during dm_scan_fdt_node(), but unfortunately
it missed adding the same logic in dm_scan_fdt_live(). This mirrors
the logic in the livetree version.
The weird thing is that commit f2006808f0 did update the test case
to test such logic, but even if I reset to that commit, the test case
still fails, and I have no idea how it could pass.
With this fix, the following 2 test cases now pass:
Test: dm_test_bus_children: bus.c
test/dm/bus.c:112, dm_test_bus_children(): num_devices ==
list_count_items(&uc->dev_head): Expected 7, got 6
Test: dm_test_fdt: test-fdt.c
test/dm/test-fdt.c:184, dm_test_fdt(): num_devices ==
list_count_items(&uc->dev_head): Expected 7, got 6
Fixes: f2006808f0 ("dm: core: parse chosen node")
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Currently the comments of several APIs (eg: dm_init_and_scan()) say:
@pre_reloc_only: If true, bind only drivers with the DM_FLAG_PRE_RELOC
flag. If false bind all drivers.
The 'Pre-Relocation Support' chapter in doc/driver-model/README.txt
documents the same that both device tree properties and driver flag
are supported.
However the implementation only checks these special device tree
properties without checking the driver flag at all. This updates
lists_bind_fdt() to consider both scenarios.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Squashed in http://patchwork.ozlabs.org/patch/996473/ :
Signed-off-by: Simon Glass <sjg@chromium.org>
