Without this patch the bcb struct could be located at an odd address
which resulted in data not being copied to the buffer.
Here was the repro steps (from Mattijs):
=> mmc dev 1
=> bcb load 1 misc
=> bcb dump command
00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
=> part start mmc 1 misc misc_start
=> mmc read ${loadaddr} ${misc_start} 4
=> bcb load 1 misc
=> bcb dump command
00000000: 62 6f 6f 74 6f 6e 63 65 2d 62 6f 6f 74 6c 6f 61
00000010: 64 65 72 00 00 00 00 00 00 00 00 00 00 00 00 00
This behavior was observed on an Amlogic A311D (ARM64) platform with a
recent GCC toolchain (11.2.0) but is most likely affecting other
platforms.
To avoid issues the structure is aligned on DMA minimum alignment value
as it is passed directly to the read function.
Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Tested-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> # on khadas vim3
On request/suggestion from Simon Glass back in May 22 2019 [1], the
'strsep' mechanism implemented in bcb_set() was set to work directly
with user-provided argv strings, to avoid duplicating memory and for
the sake of simpler implementation.
However, since we recently exposed bcb_write_reboot_reason() API to be
called by U-Boot fastboot, the idea is to be able to pass const string
literals to this new BCB API, carrying the reboot reason.
Since 'strsep' (just like its older/superseded sibling 'strtok')
modifies the input string passed as parameter, BCB command in its
current state would attempt to perform in-place modifications in a
readonly string, which might lead to unexpected results.
Fix the above with the cost of one dynamic memory allocation ('strdup').
This will also ensure no compiler warnings when passing string literals
to bcb_write_reboot_reason().
[1] http://u-boot.10912.n7.nabble.com/PATCH-v2-0-2-Add-bcb-command-to-read-modify-write-Android-BCB-td369934i20.html#a370456
Cc: Simon Glass <sjg@chromium.org>
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Fastboot is evolving and beginning with commit [1], the
upstream implementation expects bootloaders to offer support for:
- reboot-recovery
- reboot-fastboot
The most natural way to achieve the above is through a set of
pre-defined "reboot reason" strings, written into / read from
the BCB "command" field, e.g.:
- bootonce-bootloader [2]
- boot-fastboot [3]
- boot-recovery [4]
Expose the first 'bcb' API meant to be called by e.g. fastboot stack,
to allow updating the BCB reboot reason via the BCB 'command' field.
[1] https://android.googlesource.com/platform/system/core/+/dea91b4b5354af2
("Add fastbootd.")
[2] https://android.googlesource.com/platform/bootable/recovery/+/cba7fa88d8b9
("Add 'reboot bootloader' to bootloader_message.")
[3] https://android.googlesource.com/platform/bootable/recovery/+/eee4e260f9f6
("recovery: Add "boot-fastboot" command to BCB.")
[4] https://android.googlesource.com/platform/system/core/+/5e98b633a748695f
("init: Write the reason in BCB on "reboot recovery"")
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
[lukma - added missing #include <linux/errno.h> to avoid build breaks]
Enriching the functionality of U-Boot 'bcb' may assume using the
existing sub-commands as building blocks for the next ones.
A clean way to achive the above is to expose a number of static
routines, each mapped to an existing user command (e.g. load/set/store),
with a user/caller-friendly prototype (i.e. do not force the caller
to wrap an integer into a string).
This third patch makes '__bcb_store' available for internal needs.
No functional change intended.
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Enriching the functionality of U-Boot 'bcb' may assume using the
existing sub-commands as building blocks for the next ones.
A clean way to achive the above is to expose a number of static
routines, each mapped to an existing user command (e.g. load/set/store),
with a user/caller-friendly prototype (i.e. do not force the caller
to wrap an integer into a string).
This second patch makes '__bcb_set' available for internal needs.
No functional change intended.
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
Enriching the functionality of U-Boot 'bcb' may assume using the
existing sub-commands as building blocks for the next ones.
A clean way to achive the above is to expose a number of static
routines, each mapped to an existing user command (e.g. load/set/store),
with a user/caller-friendly prototype (i.e. do not force the caller
to wrap an integer into a string).
This first patch makes '__bcb_load' available for internal needs.
No functional change, except for a tiny update in error handling.
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Signed-off-by: Roman Kovalivskyi <roman.kovalivskyi@globallogic.com>
We should not use typedefs in U-Boot. They cannot be used as forward
declarations which means that header files must include the full header to
access them.
Drop the typedef and rename the struct to remove the _s suffix which is
now not useful.
This requires quite a few header-file additions.
Signed-off-by: Simon Glass <sjg@chromium.org>
We should not be using typedefs and these make it harder to use
forward declarations (to reduce header file inclusions). Drop the typedef.
Signed-off-by: Simon Glass <sjg@chromium.org>
These have been reported by Simon in [1] and fixed in [2].
However, since [1] has already been pushed to u-boot/master, the
improvements incorporated in [2] are now extracted and resubmitted.
The changes are in the area of coding style and best practices:
* s/field/fieldp/, s/size/sizep/, to convey that the variables return
an output to the caller
* s/err_1/err_read_fail/, s/err_2/err_too_small/, to be more descriptive
* Made sure 'static int do_bcb_load' appears on the same line
* Placed a `/*` on top of multi-line comment
[1] https://patchwork.ozlabs.org/patch/1104244/#2200259
[2] https://patchwork.ozlabs.org/cover/1128661/
("[v4,0/4] Add 'bcb' command to read/modify/write Android BCB")
Fixes: db7b7a05b2 ("cmd: Add 'bcb' command to read/modify/write BCB fields")
Reported-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Quote from https://patchwork.ozlabs.org/patch/1104244/#2210814:
----------8<-----------
strncmp() is chosen for the sake of paranoid/defensive programming.
Indeed, strncmp() is not really needed when comparing a variable
with a string literal. We expect strcmp() to behave safely even if the
string variable is not NUL-terminated.
In the same scenario, Linux v5.2-rc7 uses both strcmp() and strncmp(),
but the frequency of strcmp() is higher:
$ git --version
git version 2.22.0
$ (Linux 5.2-rc7) git grep -En 'strncmp\([^"]*"[[:alnum:]]+"' | wc -l
1066
$ (Linux 5.2-rc7) git grep -En 'strcmp\([^"]*"[[:alnum:]]+"' | wc -l
1968
A quick "strcmp vs strncmp" object size test shows that strcmp()
generates smaller memory footprint (gcc-8, x86_64):
$ (U-Boot) size cmd/bcb-strncmp.o cmd/bcb-strcmp.o
text data bss dec hex filename
3373 400 2048 5821 16bd cmd/bcb-strncmp.o
3314 400 2048 5762 1682 cmd/bcb-strcmp.o
So, overall, I agree to use strcmp() whenever variables are compared
with string literals.
----------8<-----------
Fixes: db7b7a05b2 ("cmd: Add 'bcb' command to read/modify/write BCB fields")
Reported-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Reviewed-by: Igor Opaniuk <igor.opaniuk@gmail.com>
Fix warning V1037 reported by PVS-Studio Static Analyzer:
Two or more case-branches perform the same actions. Check lines: 49, 53
Fixes: db7b7a05b2 ("cmd: Add 'bcb' command to read/modify/write BCB fields")
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
Reviewed-by: Igor Opaniuk <igor.opaniuk@gmail.com>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
'Bootloader Control Block' (BCB) is a well established term/acronym in
the Android namespace which refers to a location in a dedicated raw
(i.e. FS-unaware) flash (e.g. eMMC) partition, usually called "misc",
which is used as media for exchanging messages between Android userspace
(particularly recovery [1]) and an Android-capable bootloader.
On higher level, this allows implementing a subset of Android Bootloader
Requirements [2], amongst which is the Android-specific bootloader
flow [3]. Regardless how the latter is implemented in U-Boot ([3] being
the most memorable example), reading/writing/dumping the BCB fields in
the development process from inside the U-Boot is a convenient feature.
Hence, make it available to the users.
Some usage examples of the new command recorded on R-Car H3ULCB-KF
('>>>' is an overlay on top of the original console output):
=> bcb
bcb - Load/set/clear/test/dump/store Android BCB fields
Usage:
bcb load <dev> <part> - load BCB from mmc <dev>:<part>
bcb set <field> <val> - set BCB <field> to <val>
bcb clear [<field>] - clear BCB <field> or all fields
bcb test <field> <op> <val> - test BCB <field> against <val>
bcb dump <field> - dump BCB <field>
bcb store - store BCB back to mmc
Legend:
<dev> - MMC device index containing the BCB partition
<part> - MMC partition index or name containing the BCB
<field> - one of {command,status,recovery,stage,reserved}
<op> - the binary operator used in 'bcb test':
'=' returns true if <val> matches the string stored in <field>
'~' returns true if <val> matches a subset of <field>'s string
<val> - string/text provided as input to bcb {set,test}
NOTE: any ':' character in <val> will be replaced by line feed
during 'bcb set' and used as separator by upper layers
=> bcb dump command
Error: Please, load BCB first!
>>> Users must specify mmc device and partition before any other call
=> bcb load 1 misc
=> bcb load 1 1
>>> The two calls are equivalent (assuming "misc" has index 1)
=> bcb dump command
00000000: 62 6f 6f 74 6f 6e 63 65 2d 73 68 65 6c 6c 00 72 bootonce-shell.r
00000010: 79 00 72 00 00 00 00 00 00 00 00 00 00 00 00 00 y.r.............
>>> The output is in binary/string format for convenience
>>> The output size matches the size of inspected BCB field
>>> (32 bytes in case of 'command')
=> bcb test command = bootonce-shell && echo true
true
=> bcb test command = bootonce-shell- && echo true
=> bcb test command = bootonce-shel && echo true
>>> The '=' operator returns 'true' on perfect match
=> bcb test command ~ bootonce-shel && echo true
true
=> bcb test command ~ bootonce-shell && echo true
true
>>> The '~' operator returns 'true' on substring match
=> bcb set command recovery
=> bcb dump command
00000000: 72 65 63 6f 76 65 72 79 00 73 68 65 6c 6c 00 72 recovery.shell.r
00000010: 79 00 72 00 00 00 00 00 00 00 00 00 00 00 00 00 y.r.............
>>> The new value is NULL-terminated and stored in the BCB field
=> bcb set recovery "msg1:msg2:msg3"
=> bcb dump recovery
00000040: 6d 73 67 31 0a 6d 73 67 32 0a 6d 73 67 33 00 00 msg1.msg2.msg3..
00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>>> --- snip ---
>>> Every ':' is replaced by line-feed '\n' (0xA). The latter is used
>>> as separator between individual commands by Android userspace
=> bcb store
>>> Flush/store the BCB structure to MMC
[1] https://android.googlesource.com/platform/bootable/recovery
[2] https://source.android.com/devices/bootloader
[3] https://patchwork.ozlabs.org/patch/746835/
("[U-Boot,5/6] Initial support for the Android Bootloader flow")
Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>
