Please pull u-boot-marvell/master
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
---------------------------------------------------------------- - kwboot / kwbimage improvements reducing image size (Pali & Marek) - a37xx: pinctrl: Correct PWM pins definitions (Marek) - Convert the Dreamplug Ethernet and SATA to Driver Model (Tony) - serial: a37xx: DEBUG_UART improvements / fixes (Pali) ----------------------------------------------------------------
Here the Azure build, without any issues:
https://dev.azure.com/sr0718/u-boot/_build/results?buildId=104&view=resu...
Thanks, Stefan
The following changes since commit 15f7e0dc01d8a851fb1bfbf0e47eab5b67ed26b3:
Merge branch '2021-07-28-assorted-fixes' (2021-07-29 12:10:55 -0400)
are available in the Git repository at:
git@source.denx.de:u-boot/custodians/u-boot-marvell.git
for you to fetch changes up to 2cc4be280c259ac1e8e74c643c72e0331266056f:
serial: a37xx: Do not call get_ref_clk() in _debug_uart_init() (2021-07-31 09:59:59 +0200)
---------------------------------------------------------------- Marek Behún (3): tools: kwbimage: Add constant for SDIO bootfrom tools: kwbimage: Cosmetic fix - remove redundant space character arm64: a37xx: pinctrl: Correct PWM pins definitions
Pali Rohár (41): tools: kwbimage: Fix compilation without CONFIG_SYS_U_BOOT_OFFS tools: kwbimage: Simplify aligning and calculating checksum tools: kwbimage: Align SPI and NAND images to 256 bytes tools: kwbimage: Fix generation of SATA, SDIO and PCIe images tools: kwbimage: Don't crash when binary file name does not contain '/' tools: kwbimage: Fix check for v0 extended header checksum tools: kwbimage: Validate extended headers of v1 images tools: kwbimage: Validate data checksum of v1 images tools: kwbimage: Print size of binary header in kwbimage_print_header() tools: kwbimage: Use -a parameter (load address) for v1 images tools: kwbimage: Change maximum number of arguments in binary header to 256 tools: kwbimage: Fix calculating size of binary header tools: kwbimage: Add support for more BINARY headers tools: kwbimage: Don't parse PAYLOAD keyword tools: kwbimage: Add support for DATA command also for v1 images tools: kwbimage: Add support for a new DATA_DELAY command tools: kwbimage: Do not hide usage of secure header under CONFIG_ARMADA_38X tools: kwbimage: Mark all BootROM structures __packed tools: dumpimage: Fix crashing when trying to extract data from kwbimage tools: dumpimage: Show error message when trying to extract data from kwbimage tools: kwboot: Fix wrong parameter passed to read() tools: kwboot: Fix restoring terminal tools: kwboot: Print trailing newline after terminal is terminated tools: kwboot: Cosmetic fix - add missing curly brackets tools: kwboot: Check for v1 header size tools: kwboot: Fix checking image header version arm: mvebu: Fix return_to_bootrom() arm: mvebu: Mark return_to_bootrom() as a noreturn function arm: mvebu: Implement return_to_bootrom() via U-Boot's SPL framework arm: mvebu: Use U-Boot's SPL BootROM framework for booting from NAND/UART SPL: Add support for specifying offset between header and image SPL: Add support for parsing board / BootROM specific image types arm: mvebu: Load U-Boot proper binary in SPL code based on kwbimage header arm: mvebu: Remove legacy U-Boot header from kwbimage v1 files tools: kwbimage: Remove v1 kwbimage SPL padding to CONFIG_SYS_U_BOOT_OFFS bytes arm: mvebu: Remove unused macro CONFIG_SYS_U_BOOT_OFFS arm: mvebu: gdsys: Remove custom spl_board_init() kwbimage: Add support for extracting images via dumpimage tool kwbimage: Update help message about how to extract from an existing image serial: a37xx: Use CONFIG_BAUDRATE for initializing early debug UART serial: a37xx: Do not call get_ref_clk() in _debug_uart_init()
Tony Dinh (3): arm: kirkwood: Dreamplug: Add DM Ethernet and DM SATA configs arm: kirkwood: Dreamplug: Use Ethernet PHY name and address from device tree arm: kirkwood: Dreamplug: Add DM SATA and remove IDE configs
Makefile | 2 +- arch/arm/mach-mvebu/Kconfig | 26 +- arch/arm/mach-mvebu/include/mach/cpu.h | 2 +- arch/arm/mach-mvebu/lowlevel_spl.S | 3 +- arch/arm/mach-mvebu/spl.c | 248 ++++++++++-- board/Marvell/dreamplug/dreamplug.c | 62 ++- board/gdsys/a38x/Makefile | 2 +- board/gdsys/a38x/spl.c | 20 - board/kobol/helios4/Kconfig | 5 - board/solidrun/clearfog/Kconfig | 5 - common/spl/Kconfig | 4 +- common/spl/spl.c | 9 + common/spl/spl_mmc.c | 16 +- common/spl/spl_sata.c | 12 +- common/spl/spl_spi.c | 2 +- configs/clearfog_defconfig | 1 - configs/controlcenterdc_defconfig | 2 - configs/db-88f6720_defconfig | 1 - configs/db-88f6820-amc_defconfig | 1 - configs/db-88f6820-gp_defconfig | 2 - configs/db-mv784mp-gp_defconfig | 1 - configs/dreamplug_defconfig | 4 + configs/ds414_defconfig | 1 - configs/helios4_defconfig | 1 - configs/maxbcm_defconfig | 1 - configs/theadorable_debug_defconfig | 1 - configs/turris_omnia_defconfig | 1 - configs/x530_defconfig | 1 - .../pinctrl/marvell,armada-37xx-pinctrl.txt | 8 +- drivers/pinctrl/mvebu/pinctrl-armada-37xx.c | 17 +- drivers/serial/serial_mvebu_a3700.c | 9 +- include/configs/clearfog.h | 7 +- include/configs/controlcenterdc.h | 9 - include/configs/db-88f6720.h | 3 - include/configs/db-88f6820-amc.h | 5 - include/configs/db-88f6820-gp.h | 7 - include/configs/db-mv784mp-gp.h | 3 - include/configs/dreamplug.h | 12 +- include/configs/ds414.h | 5 - include/configs/helios4.h | 7 +- include/configs/theadorable.h | 3 - include/configs/turris_omnia.h | 7 - include/configs/x530.h | 3 - include/spl.h | 1 + scripts/config_whitelist.txt | 1 - tools/Makefile | 8 - tools/kwbimage.c | 437 +++++++++++++++------ tools/kwbimage.h | 54 ++- tools/kwboot.c | 24 +- 49 files changed, 730 insertions(+), 336 deletions(-) delete mode 100644 board/gdsys/a38x/spl.c
On Sat, Jul 31, 2021 at 12:04:01PM +0200, Stefan Roese wrote:
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
First off, I've applied the whole series to u-boot/master and pushed.
Second, I see from: commit 5fce2875569d6e859443af7af3477c3aebfee383 Author: Pali Rohár pali@kernel.org Date: Fri Jul 23 11:14:27 2021 +0200
SPL: Add support for specifying offset between header and image
That a number of boards are now doing: variscite_dart6ul: spl/u-boot-spl:all +144 spl/u-boot-spl:text +144 spl-u-boot-spl: add: 3/0, grow: 2/-1 bytes: 142/-4 (138) function old new delta memmove - 42 +42 spl_mmc_load 320 356 +36 __aeabi_uidivmod - 24 +24 __aeabi_idivmod - 24 +24 spl_parse_image_header 24 40 +16 board_init_r 220 216 -4
Which I think is because we need to use do_div and so rather than '/' and '%' directly in the code. Thanks!
On 01.08.21 05:28, Tom Rini wrote:
On Sat, Jul 31, 2021 at 12:04:01PM +0200, Stefan Roese wrote:
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
First off, I've applied the whole series to u-boot/master and pushed.
Second, I see from: commit 5fce2875569d6e859443af7af3477c3aebfee383 Author: Pali Rohár pali@kernel.org Date: Fri Jul 23 11:14:27 2021 +0200
SPL: Add support for specifying offset between header and imageThat a number of boards are now doing: variscite_dart6ul: spl/u-boot-spl:all +144 spl/u-boot-spl:text +144 spl-u-boot-spl: add: 3/0, grow: 2/-1 bytes: 142/-4 (138) function old new delta memmove - 42 +42 spl_mmc_load 320 356 +36 __aeabi_uidivmod - 24 +24 __aeabi_idivmod - 24 +24 spl_parse_image_header 24 40 +16 board_init_r 220 216 -4
Which I think is because we need to use do_div and so rather than '/' and '%' directly in the code. Thanks!
Pali, could you please take a look at this?
Thanks, Stefan
On Sunday 01 August 2021 07:26:51 Stefan Roese wrote:
On 01.08.21 05:28, Tom Rini wrote:
On Sat, Jul 31, 2021 at 12:04:01PM +0200, Stefan Roese wrote:
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
First off, I've applied the whole series to u-boot/master and pushed.
Second, I see from: commit 5fce2875569d6e859443af7af3477c3aebfee383 Author: Pali Rohár pali@kernel.org Date: Fri Jul 23 11:14:27 2021 +0200
SPL: Add support for specifying offset between header and imageThat a number of boards are now doing: variscite_dart6ul: spl/u-boot-spl:all +144 spl/u-boot-spl:text +144 spl-u-boot-spl: add: 3/0, grow: 2/-1 bytes: 142/-4 (138) function old new delta memmove - 42 +42 spl_mmc_load 320 356 +36 __aeabi_uidivmod - 24 +24 __aeabi_idivmod - 24 +24 spl_parse_image_header 24 40 +16 board_init_r 220 216 -4
Which I think is because we need to use do_div and so rather than '/' and '%' directly in the code. Thanks!
Pali, could you please take a look at this?
And what we can do here? 32-bit arm does not have 32-bit division instruction, so it is needed to use some sort of *idiv* function.
do_div() is macro which is doing 64-bit division by using 32-bit C operations '/' and '%', therefore it does not help with anything as this code is doing 32-bit math (not 64-bit).
Moreover in do_div() implementation is already check that first passed argument is of 64-bit type, so we cannot use it for 32-bit values.
Also note that in files which are touched by this commit are already used 32-bit division operations via C '/' operator.
So I really do not know what is expected to do here...
Thanks, Stefan
On Sun, Aug 01, 2021 at 12:46:49PM +0200, Pali Rohár wrote:
On Sunday 01 August 2021 07:26:51 Stefan Roese wrote:
On 01.08.21 05:28, Tom Rini wrote:
On Sat, Jul 31, 2021 at 12:04:01PM +0200, Stefan Roese wrote:
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
First off, I've applied the whole series to u-boot/master and pushed.
Second, I see from: commit 5fce2875569d6e859443af7af3477c3aebfee383 Author: Pali Rohár pali@kernel.org Date: Fri Jul 23 11:14:27 2021 +0200
SPL: Add support for specifying offset between header and imageThat a number of boards are now doing: variscite_dart6ul: spl/u-boot-spl:all +144 spl/u-boot-spl:text +144 spl-u-boot-spl: add: 3/0, grow: 2/-1 bytes: 142/-4 (138) function old new delta memmove - 42 +42 spl_mmc_load 320 356 +36 __aeabi_uidivmod - 24 +24 __aeabi_idivmod - 24 +24 spl_parse_image_header 24 40 +16 board_init_r 220 216 -4
Which I think is because we need to use do_div and so rather than '/' and '%' directly in the code. Thanks!
Pali, could you please take a look at this?
And what we can do here? 32-bit arm does not have 32-bit division instruction, so it is needed to use some sort of *idiv* function.
do_div() is macro which is doing 64-bit division by using 32-bit C operations '/' and '%', therefore it does not help with anything as this code is doing 32-bit math (not 64-bit).
Moreover in do_div() implementation is already check that first passed argument is of 64-bit type, so we cannot use it for 32-bit values.
Also note that in files which are touched by this commit are already used 32-bit division operations via C '/' operator.
So I really do not know what is expected to do here...
Thanks for checking. I saw block stuff and that typically does involve a 64bit value somewhere along the way. So if the answer is: - There's no 64-bit math here, really. - There's no existing shift macros we can use instead (or that ends up being larger!) - There's no existing shift macros we just need to import from the kernel.
Then we're good.
On Sunday 01 August 2021 10:33:33 Tom Rini wrote:
On Sun, Aug 01, 2021 at 12:46:49PM +0200, Pali Rohár wrote:
On Sunday 01 August 2021 07:26:51 Stefan Roese wrote:
On 01.08.21 05:28, Tom Rini wrote:
On Sat, Jul 31, 2021 at 12:04:01PM +0200, Stefan Roese wrote:
Hi Tom,
please pull the next batch of Marvell MVEBU related patches. Here the summary log:
First off, I've applied the whole series to u-boot/master and pushed.
Second, I see from: commit 5fce2875569d6e859443af7af3477c3aebfee383 Author: Pali Rohár pali@kernel.org Date: Fri Jul 23 11:14:27 2021 +0200
SPL: Add support for specifying offset between header and imageThat a number of boards are now doing: variscite_dart6ul: spl/u-boot-spl:all +144 spl/u-boot-spl:text +144 spl-u-boot-spl: add: 3/0, grow: 2/-1 bytes: 142/-4 (138) function old new delta memmove - 42 +42 spl_mmc_load 320 356 +36 __aeabi_uidivmod - 24 +24 __aeabi_idivmod - 24 +24 spl_parse_image_header 24 40 +16 board_init_r 220 216 -4
Which I think is because we need to use do_div and so rather than '/' and '%' directly in the code. Thanks!
Pali, could you please take a look at this?
And what we can do here? 32-bit arm does not have 32-bit division instruction, so it is needed to use some sort of *idiv* function.
do_div() is macro which is doing 64-bit division by using 32-bit C operations '/' and '%', therefore it does not help with anything as this code is doing 32-bit math (not 64-bit).
Moreover in do_div() implementation is already check that first passed argument is of 64-bit type, so we cannot use it for 32-bit values.
Also note that in files which are touched by this commit are already used 32-bit division operations via C '/' operator.
So I really do not know what is expected to do here...
Thanks for checking. I saw block stuff and that typically does involve a 64bit value somewhere along the way. So if the answer is:
- There's no 64-bit math here, really.
- There's no existing shift macros we can use instead (or that ends up being larger!)
- There's no existing shift macros we just need to import from the kernel.
Then we're good.
Well, "offset" member in mentioned commit is defined as "u32 offset". So it is not 64-bit for sure.
And about shift macros... Problematic part is probably mmc->read_bl_len and stor_dev->blksz. I guess that storage block size is always power of two. Or are are some mmc / sata / usb / ... devices which have block / erase size which is not power of two?
So these two members (read_bl_len and blksz) could stored as log2() value and then code can use shifts instead of division.
But such change is big in U-Boot as it would touch whole U-Boot codebase.
Maybe mmc and storage maintainers could decide if such thing is useful and try to do it?
-- Tom
participants (3)
-
Pali Rohár -
Stefan Roese -
Tom Rini