| Message ID | 1484032131-32540-1-git-send-email-hs@denx.de |
|---|---|
| State | Changes Requested |
| Delegated to: | Tom Rini |
| Headers | show |
On Tue, Jan 10, 2017 at 08:08:51AM +0100, Heiko Schocher wrote: > commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" > breaks u-boot commandline working with long commands > sending to the board. > > Since the above patch, you have to setup the fcr register. > > For board/archs which enable OF_PLATDATA, the new field > fcr in struct ns16550_platdata is not filled with a > default value ... > > This leads in not setting up the uarts fifo, which ends > in problems, when you send long commands to u-boots > commandline. > > Detected this issue with automated tbot tests on am335x > based shc board. > > The error does not popup, if you type commands. You need > to copy&paste a long command to u-boots commandshell > (or send a long command with tbot) > > Possible boards/plattforms with problems: > ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c > ./arch/arm/mach-tegra/board.c > ./board/isee/igep00x0/igep00x0.c > ./board/overo/overo.c > ./board/quipos/cairo/cairo.c > ./board/logicpd/omap3som/omap3logic.c > ./board/logicpd/zoom1/zoom1.c > ./board/timll/devkit8000/devkit8000.c > ./board/lg/sniper/sniper.c > ./board/ti/beagle/beagle.c > ./drivers/serial/serial_rockchip.c > > This patch fixes only: > ./arch/arm/mach-omap2/am33xx/board.c > > Signed-off-by: Heiko Schocher <hs@denx.de> Good find! I didn't catch this in my setups as nothing in test.py is currently long enough to trigger this apparently (I have both an am335x evm and a beagleboard in my setup). By inspection, the right fix for all platforms would be to put the old default value in as this is what was missed in the original patch from Marek.
Hello Tom, Am 10.01.2017 um 17:40 schrieb Tom Rini: > On Tue, Jan 10, 2017 at 08:08:51AM +0100, Heiko Schocher wrote: >> commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" >> breaks u-boot commandline working with long commands >> sending to the board. >> >> Since the above patch, you have to setup the fcr register. >> >> For board/archs which enable OF_PLATDATA, the new field >> fcr in struct ns16550_platdata is not filled with a >> default value ... >> >> This leads in not setting up the uarts fifo, which ends >> in problems, when you send long commands to u-boots >> commandline. >> >> Detected this issue with automated tbot tests on am335x >> based shc board. >> >> The error does not popup, if you type commands. You need >> to copy&paste a long command to u-boots commandshell >> (or send a long command with tbot) >> >> Possible boards/plattforms with problems: >> ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c >> ./arch/arm/mach-tegra/board.c >> ./board/isee/igep00x0/igep00x0.c >> ./board/overo/overo.c >> ./board/quipos/cairo/cairo.c >> ./board/logicpd/omap3som/omap3logic.c >> ./board/logicpd/zoom1/zoom1.c >> ./board/timll/devkit8000/devkit8000.c >> ./board/lg/sniper/sniper.c >> ./board/ti/beagle/beagle.c >> ./drivers/serial/serial_rockchip.c >> >> This patch fixes only: >> ./arch/arm/mach-omap2/am33xx/board.c >> >> Signed-off-by: Heiko Schocher <hs@denx.de> > > Good find! I didn't catch this in my setups as nothing in test.py is > currently long enough to trigger this apparently (I have both an am335x > evm and a beagleboard in my setup). By inspection, the right fix for > all platforms would be to put the old default value in as this is what > was missed in the original patch from Marek. The problems are boards/archs, which create "struct ns16550_platdata" on their own ... I added the original value for fcr in ./arch/arm/mach-omap2/am33xx/board.c Do you mean, I should grep for all this places and fix them in my patch (without having the chance to test it) ? Ok, seems a valid solution ... I try to do this today. bye, Heiko
On Wed, Jan 11, 2017 at 06:38:22AM +0100, Heiko Schocher wrote: > Hello Tom, > > Am 10.01.2017 um 17:40 schrieb Tom Rini: > >On Tue, Jan 10, 2017 at 08:08:51AM +0100, Heiko Schocher wrote: > >>commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" > >>breaks u-boot commandline working with long commands > >>sending to the board. > >> > >>Since the above patch, you have to setup the fcr register. > >> > >>For board/archs which enable OF_PLATDATA, the new field > >>fcr in struct ns16550_platdata is not filled with a > >>default value ... > >> > >>This leads in not setting up the uarts fifo, which ends > >>in problems, when you send long commands to u-boots > >>commandline. > >> > >>Detected this issue with automated tbot tests on am335x > >>based shc board. > >> > >>The error does not popup, if you type commands. You need > >>to copy&paste a long command to u-boots commandshell > >>(or send a long command with tbot) > >> > >>Possible boards/plattforms with problems: > >>./arch/arm/cpu/arm926ejs/lpc32xx/devices.c > >>./arch/arm/mach-tegra/board.c > >>./board/isee/igep00x0/igep00x0.c > >>./board/overo/overo.c > >>./board/quipos/cairo/cairo.c > >>./board/logicpd/omap3som/omap3logic.c > >>./board/logicpd/zoom1/zoom1.c > >>./board/timll/devkit8000/devkit8000.c > >>./board/lg/sniper/sniper.c > >>./board/ti/beagle/beagle.c > >>./drivers/serial/serial_rockchip.c > >> > >>This patch fixes only: > >>./arch/arm/mach-omap2/am33xx/board.c > >> > >>Signed-off-by: Heiko Schocher <hs@denx.de> > > > >Good find! I didn't catch this in my setups as nothing in test.py is > >currently long enough to trigger this apparently (I have both an am335x > >evm and a beagleboard in my setup). By inspection, the right fix for > >all platforms would be to put the old default value in as this is what > >was missed in the original patch from Marek. > > The problems are boards/archs, which create "struct ns16550_platdata" > on their own ... I added the original value for fcr in > ./arch/arm/mach-omap2/am33xx/board.c > > Do you mean, I should grep for all this places and > fix them in my patch (without having the chance to test it) ? > > Ok, seems a valid solution ... I try to do this today. Yes, you've got it. Thanks!
On Tue, Jan 10, 2017 at 1:08 AM, Heiko Schocher <hs@denx.de> wrote: > commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" > breaks u-boot commandline working with long commands > sending to the board. > > Since the above patch, you have to setup the fcr register. > > For board/archs which enable OF_PLATDATA, the new field > fcr in struct ns16550_platdata is not filled with a > default value ... > > This leads in not setting up the uarts fifo, which ends > in problems, when you send long commands to u-boots > commandline. > > Detected this issue with automated tbot tests on am335x > based shc board. > > The error does not popup, if you type commands. You need > to copy&paste a long command to u-boots commandshell > (or send a long command with tbot) > > Possible boards/plattforms with problems: > ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c > ./arch/arm/mach-tegra/board.c > ./board/isee/igep00x0/igep00x0.c > ./board/overo/overo.c > ./board/quipos/cairo/cairo.c > ./board/logicpd/omap3som/omap3logic.c > ./board/logicpd/zoom1/zoom1.c > ./board/timll/devkit8000/devkit8000.c > ./board/lg/sniper/sniper.c > ./board/ti/beagle/beagle.c > ./drivers/serial/serial_rockchip.c > > This patch fixes only: > ./arch/arm/mach-omap2/am33xx/board.c > > Signed-off-by: Heiko Schocher <hs@denx.de> The omap3logic is also affected. It also appears as if you hold a character down while booting, it may hang the system. I applied a similar patch to omap3logic.c and it went away. Thanks for catching that! adam > -- > 2.7.4 > > _______________________________________________ > U-Boot mailing list > U-Boot@lists.denx.de > http://lists.denx.de/mailman/listinfo/u-boot
On 01/10/2017 05:40 PM, Tom Rini wrote: > On Tue, Jan 10, 2017 at 08:08:51AM +0100, Heiko Schocher wrote: >> commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" >> breaks u-boot commandline working with long commands >> sending to the board. >> >> Since the above patch, you have to setup the fcr register. >> >> For board/archs which enable OF_PLATDATA, the new field >> fcr in struct ns16550_platdata is not filled with a >> default value ... >> >> This leads in not setting up the uarts fifo, which ends >> in problems, when you send long commands to u-boots >> commandline. >> >> Detected this issue with automated tbot tests on am335x >> based shc board. >> >> The error does not popup, if you type commands. You need >> to copy&paste a long command to u-boots commandshell >> (or send a long command with tbot) >> >> Possible boards/plattforms with problems: >> ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c >> ./arch/arm/mach-tegra/board.c >> ./board/isee/igep00x0/igep00x0.c >> ./board/overo/overo.c >> ./board/quipos/cairo/cairo.c >> ./board/logicpd/omap3som/omap3logic.c >> ./board/logicpd/zoom1/zoom1.c >> ./board/timll/devkit8000/devkit8000.c >> ./board/lg/sniper/sniper.c >> ./board/ti/beagle/beagle.c >> ./drivers/serial/serial_rockchip.c >> >> This patch fixes only: >> ./arch/arm/mach-omap2/am33xx/board.c >> >> Signed-off-by: Heiko Schocher <hs@denx.de> > > Good find! I didn't catch this in my setups as nothing in test.py is > currently long enough to trigger this apparently (I have both an am335x > evm and a beagleboard in my setup). By inspection, the right fix for > all platforms would be to put the old default value in as this is what > was missed in the original patch from Marek. > It was probably missed because this patch pre-dates the OF_PLATDATA addition and was stuck in the ML for too long.
Hello Adam, Am 11.01.2017 um 15:44 schrieb Adam Ford: > On Tue, Jan 10, 2017 at 1:08 AM, Heiko Schocher <hs@denx.de> wrote: >> commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" >> breaks u-boot commandline working with long commands >> sending to the board. >> >> Since the above patch, you have to setup the fcr register. >> >> For board/archs which enable OF_PLATDATA, the new field >> fcr in struct ns16550_platdata is not filled with a >> default value ... >> >> This leads in not setting up the uarts fifo, which ends >> in problems, when you send long commands to u-boots >> commandline. >> >> Detected this issue with automated tbot tests on am335x >> based shc board. >> >> The error does not popup, if you type commands. You need >> to copy&paste a long command to u-boots commandshell >> (or send a long command with tbot) >> >> Possible boards/plattforms with problems: >> ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c >> ./arch/arm/mach-tegra/board.c >> ./board/isee/igep00x0/igep00x0.c >> ./board/overo/overo.c >> ./board/quipos/cairo/cairo.c >> ./board/logicpd/omap3som/omap3logic.c >> ./board/logicpd/zoom1/zoom1.c >> ./board/timll/devkit8000/devkit8000.c >> ./board/lg/sniper/sniper.c >> ./board/ti/beagle/beagle.c >> ./drivers/serial/serial_rockchip.c >> >> This patch fixes only: >> ./arch/arm/mach-omap2/am33xx/board.c >> >> Signed-off-by: Heiko Schocher <hs@denx.de> > > The omap3logic is also affected. It also appears as if you hold a > character down while booting, it may hang the system. I applied a > similar patch to omap3logic.c and it went away. Thanks for testing! > Thanks for catching that! Thanks goes to tbot [1] and weekly automated u-boot tests with tbot integrated into buildbot [2] (I also automated "git bisect" to find the, which commit breaks a board ... if interested in a demo video, see[5]) bye, Heiko [1] https://github.com/hsdenx/tbot [2] http://xeidos.ddns.net/buildbot/tgrid [3] tbot result webpage http://xeidos.ddns.net/tests/test_db_auslesen.php [4] tbot dokumentation http://www.tbot.tools/main.html [5] tbot demo on youtube https://youtu.be/zfjpj3DLsx4 Its uncutted, so very long, but in the video comment I added timemarkers, where the interesting parts are ...
Hi Marek, Am 11.01.2017 um 16:11 schrieb Marek Vasut: > On 01/10/2017 05:40 PM, Tom Rini wrote: >> On Tue, Jan 10, 2017 at 08:08:51AM +0100, Heiko Schocher wrote: >>> commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" >>> breaks u-boot commandline working with long commands >>> sending to the board. >>> >>> Since the above patch, you have to setup the fcr register. >>> >>> For board/archs which enable OF_PLATDATA, the new field >>> fcr in struct ns16550_platdata is not filled with a >>> default value ... >>> >>> This leads in not setting up the uarts fifo, which ends >>> in problems, when you send long commands to u-boots >>> commandline. >>> >>> Detected this issue with automated tbot tests on am335x >>> based shc board. >>> >>> The error does not popup, if you type commands. You need >>> to copy&paste a long command to u-boots commandshell >>> (or send a long command with tbot) >>> >>> Possible boards/plattforms with problems: >>> ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c >>> ./arch/arm/mach-tegra/board.c >>> ./board/isee/igep00x0/igep00x0.c >>> ./board/overo/overo.c >>> ./board/quipos/cairo/cairo.c >>> ./board/logicpd/omap3som/omap3logic.c >>> ./board/logicpd/zoom1/zoom1.c >>> ./board/timll/devkit8000/devkit8000.c >>> ./board/lg/sniper/sniper.c >>> ./board/ti/beagle/beagle.c >>> ./drivers/serial/serial_rockchip.c >>> >>> This patch fixes only: >>> ./arch/arm/mach-omap2/am33xx/board.c >>> >>> Signed-off-by: Heiko Schocher <hs@denx.de> >> >> Good find! I didn't catch this in my setups as nothing in test.py is >> currently long enough to trigger this apparently (I have both an am335x >> evm and a beagleboard in my setup). By inspection, the right fix for >> all platforms would be to put the old default value in as this is what >> was missed in the original patch from Marek. >> > It was probably missed because this patch pre-dates the OF_PLATDATA > addition and was stuck in the ML for too long. No problem, I prepare a fix. But it shows how important would be, that we have as much as possible boards in a real life test setup... (and of course, as much as possible testcases) bye, Heiko
diff --git a/arch/arm/mach-omap2/am33xx/board.c b/arch/arm/mach-omap2/am33xx/board.c index 581c0ab..a0ce62c 100644 --- a/arch/arm/mach-omap2/am33xx/board.c +++ b/arch/arm/mach-omap2/am33xx/board.c @@ -39,15 +39,26 @@ DECLARE_GLOBAL_DATA_PTR; #if !CONFIG_IS_ENABLED(OF_CONTROL) +/* Clear & enable FIFOs */ +#define UART_FCRVAL (UART_FCR_FIFO_EN | \ + UART_FCR_RXSR | \ + UART_FCR_TXSR) + static const struct ns16550_platdata am33xx_serial[] = { - { .base = CONFIG_SYS_NS16550_COM1, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, + { .base = CONFIG_SYS_NS16550_COM1, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, # ifdef CONFIG_SYS_NS16550_COM2 - { .base = CONFIG_SYS_NS16550_COM2, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, + { .base = CONFIG_SYS_NS16550_COM2, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, # ifdef CONFIG_SYS_NS16550_COM3 - { .base = CONFIG_SYS_NS16550_COM3, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, - { .base = CONFIG_SYS_NS16550_COM4, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, - { .base = CONFIG_SYS_NS16550_COM5, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, - { .base = CONFIG_SYS_NS16550_COM6, .reg_shift = 2, .clock = CONFIG_SYS_NS16550_CLK }, + { .base = CONFIG_SYS_NS16550_COM3, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, + { .base = CONFIG_SYS_NS16550_COM4, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, + { .base = CONFIG_SYS_NS16550_COM5, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, + { .base = CONFIG_SYS_NS16550_COM6, .reg_shift = 2, + .clock = CONFIG_SYS_NS16550_CLK, .fcr = UART_FCRVAL }, # endif # endif };
commit: 65f83802b7a5b "serial: 16550: Add getfcr accessor" breaks u-boot commandline working with long commands sending to the board. Since the above patch, you have to setup the fcr register. For board/archs which enable OF_PLATDATA, the new field fcr in struct ns16550_platdata is not filled with a default value ... This leads in not setting up the uarts fifo, which ends in problems, when you send long commands to u-boots commandline. Detected this issue with automated tbot tests on am335x based shc board. The error does not popup, if you type commands. You need to copy&paste a long command to u-boots commandshell (or send a long command with tbot) Possible boards/plattforms with problems: ./arch/arm/cpu/arm926ejs/lpc32xx/devices.c ./arch/arm/mach-tegra/board.c ./board/isee/igep00x0/igep00x0.c ./board/overo/overo.c ./board/quipos/cairo/cairo.c ./board/logicpd/omap3som/omap3logic.c ./board/logicpd/zoom1/zoom1.c ./board/timll/devkit8000/devkit8000.c ./board/lg/sniper/sniper.c ./board/ti/beagle/beagle.c ./drivers/serial/serial_rockchip.c This patch fixes only: ./arch/arm/mach-omap2/am33xx/board.c Signed-off-by: Heiko Schocher <hs@denx.de> --- arch/arm/mach-omap2/am33xx/board.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)