Re: how do i add the defn for the beagle xM to barebox?
Le Sun, 5 Feb 2012 16:04:07 -0500 (EST), Robert P. J. Day rpj...@crashcourse.ca a écrit : but this isn't a small difference -- this represents potentially an entire subsystem (NAND) that i want the ability to de-activate. in a perfect world, i want the ability to say concisely that i have a beagle but i have no NAND flash, period. i don't want any of the NAND code compiled, i don't want it included in the binary, so obviously i don't want any of that code to run, only to realize it has nothing to do. well it obvious that is you don't want the NAND code to be compiled in, you simply need to unselect the options which enables it ! Eric -- http://eukrea.com/en/news/104-2012 ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
On Mon, 6 Feb 2012, Sascha Hauer wrote: ... snip ... The mentioned U-Boot code does not use any of the nand layer but only some omap specific function to detect the nand id (identify_nand_chip). This means that you can add this code to barebox and still disable nand support. If you still want to have a more fine grained config we can add a config MACH_PANDA_XM_VARIANT bool config MACH_PANDA_XM_DETECT_RAM_TYPE bool But we should add such options with care. Generally I don't really like to ask the user complicated questions for things that can be autodetected. don't take me too seriously yet as i'm still wandering around the code base figuring out how things work. :-) thanks to eric for clearing up a number of things for me, i now have a much better idea of what's worth considering and what isn't. rday -- Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
Hi Robert, Le Sun, 5 Feb 2012 06:52:28 -0500 (EST), Robert P. J. Day rpj...@crashcourse.ca a écrit : thoughts? i'd actually like to make the tweaks one at a time so that the git log shows clearly how the two boards differ one feature at a time. By sensing a few GPIO, you can easily check which board you are running barebox on and thus have ony binary which supports all beagleboards (as already done in u-boot). Check around line 108 in u-boot's beagle support for the details : http://git.denx.de/?p=u-boot.git;a=blob;f=board/ti/beagle/beagle.c;h=5c04b34e1ab0140aca93e7752c8672ce7aebc0b9;hb=HEAD So in the end you can register the nand support only if you run on a board which has a NAND flash and have one defconfig for all beagleboards (some tweaks in the clock settings may be nedded - especially for USB IIRC - as beagle XM is built around a DM3730 which can run at 1GHz when older beagles had older OMAP3530 limited to 720MHz). Eric -- http://eukrea.com/en/news/104-2012 ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
On Sun, 5 Feb 2012, Eric Bénard wrote: Hi Robert, Le Sun, 5 Feb 2012 06:52:28 -0500 (EST), Robert P. J. Day rpj...@crashcourse.ca a écrit : thoughts? i'd actually like to make the tweaks one at a time so that the git log shows clearly how the two boards differ one feature at a time. By sensing a few GPIO, you can easily check which board you are running barebox on and thus have ony binary which supports all beagleboards (as already done in u-boot). Check around line 108 in u-boot's beagle support for the details : http://git.denx.de/?p=u-boot.git;a=blob;f=board/ti/beagle/beagle.c;h=5c04b34e1ab0140aca93e7752c8672ce7aebc0b9;hb=HEAD So in the end you can register the nand support only if you run on a board which has a NAND flash and have one defconfig for all beagleboards (some tweaks in the clock settings may be nedded - especially for USB IIRC - as beagle XM is built around a DM3730 which can run at 1GHz when older beagles had older OMAP3530 limited to 720MHz). ah, thank you, that might be what i'm after. we'll see if that's enough to handle all of the differences. i may very well be back with more questions later. in fact, count on it. :-) rday -- Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
Le Sun, 5 Feb 2012 10:37:59 -0500 (EST),
Robert P. J. Day rpj...@crashcourse.ca a écrit :
On Sun, 5 Feb 2012, Eric Bénard wrote:
Hi Robert,
Le Sun, 5 Feb 2012 06:52:28 -0500 (EST),
Robert P. J. Day rpj...@crashcourse.ca a écrit :
thoughts? i'd actually like to make the tweaks one at a time so
that the git log shows clearly how the two boards differ one feature
at a time.
By sensing a few GPIO, you can easily check which board you are running
barebox on and thus have ony binary which supports all beagleboards
(as already done in u-boot).
Check around line 108 in u-boot's beagle support for the details :
http://git.denx.de/?p=u-boot.git;a=blob;f=board/ti/beagle/beagle.c;h=5c04b34e1ab0140aca93e7752c8672ce7aebc0b9;hb=HEAD
So in the end you can register the nand support only if you run on a
board which has a NAND flash and have one defconfig for all
beagleboards (some tweaks in the clock settings may be nedded -
especially for USB IIRC - as beagle XM is built around a DM3730 which
can run at 1GHz when older beagles had older OMAP3530 limited to
720MHz).
h ... i see one possible problem that maybe means i just haven't
read far enough. in u-boot, in that beagle.c file, here's a chunk of
code that identifies the NAND chip, of which there is none on the xM:
/*
* We need to identify what PoP memory is on the board so that
* we know what timings to use. If we can't identify it then
* we know it's an xM. To map the ID values please see nand_ids.c
*/
identify_nand_chip(pop_mfr, pop_id);
*mr = MICRON_V_MR_165;
switch (get_board_revision()) {
case REVISION_C4:
if (pop_mfr == NAND_MFR_STMICRO pop_id == 0xba) {
/* 512MB DDR */
*mcfg = NUMONYX_V_MCFG_165(512 20);
*ctrla = NUMONYX_V_ACTIMA_165;
*ctrlb = NUMONYX_V_ACTIMB_165;
*rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_165MHz;
break;
} else if (pop_mfr == NAND_MFR_MICRON pop_id == 0xbc) {
/* Beagleboard Rev C5, 256MB DDR */
*mcfg = MICRON_V_MCFG_200(256 20);
*ctrla = MICRON_V_ACTIMA_200;
*ctrlb = MICRON_V_ACTIMB_200;
*rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_200MHz;
break;
}
case REVISION_XM_A:
case REVISION_XM_B:
case REVISION_XM_C:
if (pop_mfr == 0) {
/* 256MB DDR */
*mcfg = MICRON_V_MCFG_200(256 20);
*ctrla = MICRON_V_ACTIMA_200;
*ctrlb = MICRON_V_ACTIMB_200;
*rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_200MHz;
} else {
/* 512MB DDR */
*mcfg = NUMONYX_V_MCFG_165(512 20);
*ctrla = NUMONYX_V_ACTIMA_165;
*ctrlb = NUMONYX_V_ACTIMB_165;
*rfr_ctrl = SDP_3430_SDRC_RFR_CTRL_165MHz;
}
break;
... snip ...
ok, i can see how that works, but if you need to test this way, it
means that you'll constantly be overriding the CONFIG settings in the
beagle defconfig file.
if i read you correctly, then one should be able to use the same
defconfig file for both a regular beagle and an xM. but the current
barebox beagle xload defconfig file contains stuff like:
CONFIG_MTD=y
# CONFIG_MTD_WRITE is not set
# CONFIG_MTD_OOB_DEVICE is not set
CONFIG_NAND=y
# CONFIG_NAND_ECC_SOFT is not set
# CONFIG_NAND_ECC_HW_SYNDROME is not set
# CONFIG_NAND_ECC_HW_NONE is not set
# CONFIG_NAND_INFO is not set
# CONFIG_NAND_BBT is not set
CONFIG_NAND_OMAP_GPMC=y
which is fine for a regular beagle but totally wrong for an xM. does
this mean that, with an xM, i'd include the same (incorrect) defconfig
file, and have to be constantly overriding some of those settings?
i can see how getting the board revision lets you test how to do
things differently across a related set of boards. but the difference
between a regular beagle and an xM is *sizable* when you're talking
about NAND versus no NAND. ideally, i'd like to have that board
revision test available in the defconfig file itself. :-)
This is the case where u-boot is built with SPL thus reoplacing
x-load : they always read NAND ID as some XM boards were mounted with
Numonyx POP which includes NAND and 166MHz RAM when mot XM boards have
a POP with only 200MHz DDR (and in that case manufacturer ID is 0). So
even on a XM you need to check the NAND ID to set the right RAM
settings.
Eric
--
http://eukrea.com/en/news/104-2012
___
barebox mailing list
Re: how do i add the defn for the beagle xM to barebox?
On Sun, 5 Feb 2012, Eric Bénard wrote: This is the case where u-boot is built with SPL thus reoplacing x-load : they always read NAND ID as some XM boards were mounted with Numonyx POP which includes NAND and 166MHz RAM when mot XM boards have a POP with only 200MHz DDR (and in that case manufacturer ID is 0). So even on a XM you need to check the NAND ID to set the right RAM settings. ah, got it. but what about in a more general case? what if you have a current, accurate definition for an existing board? then a *slight* variant of that board comes along, for which some settings in the defconfig file are simply wrong? are there any examples of that in barebox right now? and if not, how would one handle them? put another way, what if *all* xM boards had no NAND? then we'd be back to my original question, and it's still not clear how you'd define that new board for barebox. rday -- Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
Le Sun, 5 Feb 2012 13:00:13 -0500 (EST), Robert P. J. Day rpj...@crashcourse.ca a écrit : On Sun, 5 Feb 2012, Eric Bénard wrote: This is the case where u-boot is built with SPL thus reoplacing x-load : they always read NAND ID as some XM boards were mounted with Numonyx POP which includes NAND and 166MHz RAM when mot XM boards have a POP with only 200MHz DDR (and in that case manufacturer ID is 0). So even on a XM you need to check the NAND ID to set the right RAM settings. ah, got it. but what about in a more general case? what if you have a current, accurate definition for an existing board? then a *slight* variant of that board comes along, for which some settings in the defconfig file are simply wrong? are there any examples of that in barebox right now? and if not, how would one handle them? put another way, what if *all* xM boards had no NAND? then we'd be back to my original question, and it's still not clear how you'd define that new board for barebox. if no XM board had NAND, you would simply check the board type using the GPIO sampled and you wouldn't register the nand (line 305 in board-beagle.c) and the fact that the nand driver is enabled is not a problem if the device is not registrered the driver won't be used. Eric -- http://eukrea.com/en/news/104-2012 ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
Re: how do i add the defn for the beagle xM to barebox?
On Sun, 5 Feb 2012, Eric Bénard wrote: Le Sun, 5 Feb 2012 13:00:13 -0500 (EST), Robert P. J. Day rpj...@crashcourse.ca a écrit : On Sun, 5 Feb 2012, Eric Bénard wrote: This is the case where u-boot is built with SPL thus reoplacing x-load : they always read NAND ID as some XM boards were mounted with Numonyx POP which includes NAND and 166MHz RAM when mot XM boards have a POP with only 200MHz DDR (and in that case manufacturer ID is 0). So even on a XM you need to check the NAND ID to set the right RAM settings. ah, got it. but what about in a more general case? what if you have a current, accurate definition for an existing board? then a *slight* variant of that board comes along, for which some settings in the defconfig file are simply wrong? are there any examples of that in barebox right now? and if not, how would one handle them? put another way, what if *all* xM boards had no NAND? then we'd be back to my original question, and it's still not clear how you'd define that new board for barebox. if no XM board had NAND, you would simply check the board type using the GPIO sampled and you wouldn't register the nand (line 305 in board-beagle.c) and the fact that the nand driver is enabled is not a problem if the device is not registrered the driver won't be used. not to put too fine a point on it but ... yuck. don't get me wrong, i certainly see the value in testing things like board versions so the code knows how to handle small differences between similar boards (memory speed, flash types or sizes and so on). but this isn't a small difference -- this represents potentially an entire subsystem (NAND) that i want the ability to de-activate. in a perfect world, i want the ability to say concisely that i have a beagle but i have no NAND flash, period. i don't want any of the NAND code compiled, i don't want it included in the binary, so obviously i don't want any of that code to run, only to realize it has nothing to do. i haven't finished reading the docs, but i'm assuming there's no way to do that trivially easily. rday -- Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ___ barebox mailing list barebox@lists.infradead.org http://lists.infradead.org/mailman/listinfo/barebox
