Arjan van de Ven wrote: > On Thu, 2005-06-09 at 10:04 -0700, Tim Bird wrote: > >>I'm trying to use virt_to_phys() to find the physical address >>of some kernel RAM. (This is an on ARM OSK board). >>It's not working as I expected: > > what do you wan to do with the "physical" address? You can't use it for > letting any device do DMA to it or so for example; you HAVE to use the > dma mapping for that instead. > It's a little embarrassing to continue the thread, since I made such a stupid mistake with the printf format option. However, since you ask... I'm trying to debug a problem with the kill-printk.patch on the OSK board. With printk turned off, the serial console gets screwed up during booting. With printk turned on, the serial console is OK. I wrote a routine to output stuff to the serial port even with printk's turned off, but once the serial port is initialized by the regular driver, the interaction between my routine and the serial console causes its own set of problems. I loaned my jtag device to a colleague, there are no other serial ports on the device, and there's no video output, so essentially I'm flying blind (but only at the exact moment I need the debug output ;-). I can output stuff before the kernel loads (including dumping RAM content before the page tables are set up), so I'm trying to stuff some debug info into memory, and print it out from the decompression code (pre-kernel execution) on the next machine restart (warm). For this, I need the physical address of the RAM where the debug information was put. Does this sound sane or is there an easier way? ============================= Tim Bird Architecture Group Chair, CE Linux Forum Senior Staff Engineer, Sony Electronics ============================= -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
