-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 10/09/12 15:09, Stephen Warren wrote:
On 10/09/2012 03:38 PM, Tom Rini wrote:
On Tue, Oct 09, 2012 at 02:33:51PM -0600, Stephen Warren wrote:
On 09/29/2012 03:53 PM, Marek Vasut wrote:
Enable CONFIG_SERIAL_MULTI for all builds of U-Boot. That includes both SPL builds and non-SPL builds, everything. To avoid poluting this patch with removal of ifdef-endif constructions containing CONFIG_SERIAL_MULTI, the CONFIG_SERIAL_MULTI is temporarily added into CPPFLAGS in config.mk . This will be again removed in following patch.
Marek,
This patch (at least, the version of it checked into u-boot/next) breaks U-Boot on Tegra.
I believe the reason is because nothing ever calls serial_initialize() in the SPL on Tegra. If I edit arch/arm/cpu/arm720t/tegra20/spl.c function preloader_console_init() to call serial_initialize() right before it calls serial_init(), then everything works fine. Without this, drivers/serial/serial.c is never initialised (not even BSS cleared!) and so it's left set to some bogus value, and so get_current() doesn't replace it with default_serial_console(), so an invalid pointer is accessed, causing a hang or crash.
I'm not sure quite what the correct solution is here. Is Tegra's custom spl.c doing too much; stuff that should come from the common spl.c is cut/paste here? Or, is adding the call to serial_initialize() the correct fix? Or, is the root-cause of the problem that BSS (serial_current) isn't being cleared to 0?
Tegra SPL needs to either be updated ala common/spl/ (preloader_console_init() calls serial_init()) or switched over to it.
Hmm, well it's already calling serial_init() just not serial_initialize(). Perhaps the issue is the following code missing from Tegra's SPL:
/* Clear the BSS. */ memset(__bss_start, 0, __bss_end__ - __bss_start);
Anyway, I'll let Allen take a look at it, since he's most familiar with Tegra SPL.
Yes, if SPL wasn't previously clearing the BSS, this would be a problem now.
- -- Tom