This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [PATCHES] Update to pcmb memory sizing and two tiny fixes


Ian Campbell wrote:
> 
> Hi all,
> 
> Sometime ago I sent memsize-fixes.patch to the list which allows the
> choice at compile time of hardcoding the memory regions or probing the
> BIOS. Fortunately this patch doesn't seem to have been applied yet as it
> contains a glaring error. At some point I renamed the CDL option, but
> neglected to also rename the #ifdef in the c file. I have attached
> memsize-fixes-2.patch which is corrected. (It also changes
> CYGNUM_HAL_RTC_PERIOD to use default_value rather than calculated, which
> allows the value to be overriden in the board specific HAL)

FWIW I've changed the macro names again (only the outer component deserved
the CYGPKG prefix).
 
> In the process of getting the FLASH to work on our board I noticed that
> the macro HAL_[ID]CACHE_IS_ENABLED was not defined for i386 targets
> (preventing the flash stuff from building). The attached patch
> 'cache-is-enabled.patch' implements this in the spirit of the other i386
> CACHE macros (ie does nothing).

Yes, although if you're not using a PC board I assume you have some BIOS
that will enable the caches really as ecos won't.
 
> Finally, I tried to use the tcp-echo test in the net package, the target
> part was unable to setup it's network via DHCP, I think because the main
> thread and load threads were of higher priority then the networking
> thread, so the DHCP replies were not received. tcp-echo-test.patch
> delays the starting of the load threads until after the network is
> initialised and starts the main thread with a lower priority, bumping it
> up after the network has initialised.

This we're not sure about at all. The test is deliberately intended to have
the main and load threads at higher priority so that the net thread
deliberately only does get time to run when everything else is waiting for
packets. And the call to init_all_network_interfaces gets the DHCP lease
then and there. So unless your DHCP lease timeouts are ridiculously short
it should work.

Are you sure the system clock is correctly set up so that the DHCP timeouts
are correct? Try disconnecting the ethernet and checking it reports a
filaure to lease in the correct amount of time (5s I think).

I applied the first two. But again I'll remind you about the copyright
assignment - each individual patch has fortunately been small enough.

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]