| Message ID | 1372201560-26015-1-git-send-email-linus.walleij@linaro.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Jun 26, 2013 at 01:06:00AM +0200, Linus Walleij wrote: > Instead of remapping the IO memory very early with > pci_map_io_early(), use pci_ioremap_io() when setting > up resources as most PCI bridges do. Also request this > resource properly. What impact does this have with a VGA card plugged in?
On Wednesday 26 June 2013, Russell King - ARM Linux wrote: > On Wed, Jun 26, 2013 at 01:06:00AM +0200, Linus Walleij wrote: > > Instead of remapping the IO memory very early with > > pci_map_io_early(), use pci_ioremap_io() when setting > > up resources as most PCI bridges do. Also request this > > resource properly. > > What impact does this have with a VGA card plugged in? It probably won't work. It's my fault for misreading the driver: I saw it didn't do the pci_ioremap_io, but didn't remember the *_early call being present in the platform. Arnd
On Wed, Jun 26, 2013 at 5:12 PM, Arnd Bergmann <arnd@arndb.de> wrote: > On Wednesday 26 June 2013, Russell King - ARM Linux wrote: >> On Wed, Jun 26, 2013 at 01:06:00AM +0200, Linus Walleij wrote: >> > Instead of remapping the IO memory very early with >> > pci_map_io_early(), use pci_ioremap_io() when setting >> > up resources as most PCI bridges do. Also request this >> > resource properly. >> >> What impact does this have with a VGA card plugged in? > > It probably won't work. It's my fault for misreading the driver: > I saw it didn't do the pci_ioremap_io, but didn't remember the > *_early call being present in the platform. I haven't got the VGA working (hacking on the biosemu in U-boot from time to time, will get there I hope). So keep this patch out for now. However I don't see how the VGA console is using iospace? We set vga_base to the start of non-prefetched memory, and assume the first slot is the VGA card, then it goes poking into the screen memory or something... but maybe it's poking the ioports in some runpath I can't see :-P Yours, Linus Walleij
On Wednesday 26 June 2013, Linus Walleij wrote: > On Wed, Jun 26, 2013 at 5:12 PM, Arnd Bergmann <arnd@arndb.de> wrote: > > On Wednesday 26 June 2013, Russell King - ARM Linux wrote: > >> On Wed, Jun 26, 2013 at 01:06:00AM +0200, Linus Walleij wrote: > >> > Instead of remapping the IO memory very early with > >> > pci_map_io_early(), use pci_ioremap_io() when setting > >> > up resources as most PCI bridges do. Also request this > >> > resource properly. > >> > >> What impact does this have with a VGA card plugged in? > > > > It probably won't work. It's my fault for misreading the driver: > > I saw it didn't do the pci_ioremap_io, but didn't remember the > > *_early call being present in the platform. > > I haven't got the VGA working (hacking on the biosemu in > U-boot from time to time, will get there I hope). So keep this > patch out for now. > > However I don't see how the VGA console is using iospace? > We set vga_base to the start of non-prefetched memory, and > assume the first slot is the VGA card, then it goes poking into > the screen memory or something... but maybe it's poking the > ioports in some runpath I can't see :-P > I think you are looking for vga_io_w()/vga_io_r() in case of vga16fb. vgacon uses inb_p/outb_p directly. Arnd
diff --git a/arch/arm/mach-integrator/pci_v3.c b/arch/arm/mach-integrator/pci_v3.c index bef1005..029d670 100644 --- a/arch/arm/mach-integrator/pci_v3.c +++ b/arch/arm/mach-integrator/pci_v3.c @@ -538,6 +538,13 @@ static int __init pci_v3_setup_resources(struct pci_sys_data *sys) "memory region\n"); return -EBUSY; } + if (request_resource(&iomem_resource, &io_mem)) { + release_resource(&io_mem); + printk(KERN_ERR "PCI: unable to allocate I/O " + "memory region\n"); + return -EBUSY; + } + pci_ioremap_io(0, io_mem.start); /* * the mem resource for this bus @@ -1039,6 +1046,5 @@ int __init pci_v3_early_init(void) { iotable_init(pci_v3_io_desc, ARRAY_SIZE(pci_v3_io_desc)); vga_base = (unsigned long)PCI_MEMORY_VADDR; - pci_map_io_early(__phys_to_pfn(PHYS_PCI_IO_BASE)); return 0; }
Instead of remapping the IO memory very early with pci_map_io_early(), use pci_ioremap_io() when setting up resources as most PCI bridges do. Also request this resource properly. Reported-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Linus Walleij <linus.walleij@linaro.org> --- arch/arm/mach-integrator/pci_v3.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)