Re: PR/48431 CVS commit: src/sys/arch/alpha/pci

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

The following reply was made to PR port-alpha/48431; it has been noted by GNATS.

From: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: nullnilaki%gmail.com@localhost, tsutsui%ceres.dti.ne.jp@localhost
Subject: Re: PR/48431 CVS commit: src/sys/arch/alpha/pci
Date: Sun, 19 Jan 2014 12:38:00 +0900

 >  >  --- dec_6600.c 13 Oct 2012 17:58:54 -0000      1.34
 >  >  +++ dec_6600.c 13 Jan 2014 16:18:09 -0000
 >  >  @@ -201,6 +201,21 @@ dec_6600_device_register(device_t dev, v
 >  >         struct bootdev_data *b = bootdev_data;
 >  >         device_t parent = device_parent(dev);
 >  >
 >  >  +      /* set properties for PCI framebuffers */
  :
 
 >  I apply this patch.
 >  It work fine and radeonfb attached correctly.
 
 Thanks.
 
 After some thoughts, I notice x86 ports have device_pci_register()
 in sys/arch/x86/pci/pci_machidep.c and now I think it's better to put
 a common PCI specific device_register function in alpha/pci/pci_machidep.c.
 
 Could you also try the following patch (after revert previous
 dec_6600.c diff)?
 
 ---
 Index: alpha/autoconf.c
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/alpha/alpha/autoconf.c,v
 retrieving revision 1.52
 diff -u -p -r1.52 autoconf.c
 --- alpha/autoconf.c   29 Jul 2012 18:05:39 -0000      1.52
 +++ alpha/autoconf.c   18 Jan 2014 14:08:35 -0000
 @@ -44,6 +44,8 @@
  
  __KERNEL_RCSID(0, "$NetBSD: autoconf.c,v 1.52 2012/07/29 18:05:39 mlelstv Exp 
$");
  
 +#include "pci.h"
 +
  #include <sys/param.h>
  #include <sys/systm.h>
  #include <sys/buf.h>
 @@ -53,6 +55,8 @@ __KERNEL_RCSID(0, "$NetBSD: autoconf.c,v
  #include <sys/conf.h>
  #include <dev/cons.h>
  
 +#include <dev/pci/pcivar.h>
 +
  #include <machine/autoconf.h>
  #include <machine/alpha.h>
  #include <machine/cpu.h>
 @@ -172,6 +176,13 @@ atoi(const char *s)
  void
  device_register(device_t dev, void *aux)
  {
 +#if NPCI > 0
 +      device_t parent = device_parent(dev);
 +
 +      if (parent != NULL && device_is_a(parent, "pci"))
 +              device_pci_register(dev, aux);
 +#endif
 +
        if (bootdev_data == NULL) {
                /*
                 * There is no hope.
 Index: include/pci_machdep.h
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/alpha/include/pci_machdep.h,v
 retrieving revision 1.15
 diff -u -p -r1.15 pci_machdep.h
 --- include/pci_machdep.h      6 Feb 2012 02:14:13 -0000       1.15
 +++ include/pci_machdep.h      18 Jan 2014 14:08:35 -0000
 @@ -109,3 +109,4 @@ void       pci_display_console(bus_space_tag_t
      ((c)->pc_pciide_compat_intr_establish == NULL ? NULL :            \
       (*(c)->pc_pciide_compat_intr_establish)((c)->pc_conf_v, (d), (p),        
\
        (ch), (f), (a)))
 +void  device_pci_register(device_t, void *);
 Index: pci/pci_machdep.c
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/alpha/pci/pci_machdep.c,v
 retrieving revision 1.19
 diff -u -p -r1.19 pci_machdep.c
 --- pci/pci_machdep.c  6 Feb 2012 02:14:15 -0000       1.19
 +++ pci/pci_machdep.c  18 Jan 2014 14:08:35 -0000
 @@ -59,6 +59,8 @@ __KERNEL_RCSID(0, "$NetBSD: pci_machdep.
  #include <dev/pci/tgavar.h>
  #endif
  
 +#include <machine/rpb.h>
 +
  void
  pci_display_console(bus_space_tag_t iot, bus_space_tag_t memt, 
pci_chipset_tag_t pc, int bus, int device, int function)
  {
 @@ -101,3 +103,20 @@ pci_display_console(bus_space_tag_t iot,
                panic("pci_display_console: unconfigured device at %d/%d/%d",
                    bus, device, function);
  }
 +
 +void
 +device_pci_register(device_t dev, void *aux)
 +{
 +      struct pci_attach_args *pa = aux;
 +      struct ctb *ctb;
 +      prop_dictionary_t dict;
 +
 +      /* set properties for PCI framebuffers */
 +      ctb = (struct ctb *)(((char *)hwrpb) + hwrpb->rpb_ctb_off);
 +      if (PCI_CLASS(pa->pa_class) == PCI_CLASS_DISPLAY &&
 +          ctb->ctb_term_type == CTB_GRAPHICS) {
 +              /* XXX should consider multiple displays? */
 +              dict = device_properties(dev);
 +              prop_dictionary_set_bool(dict, "is_console", true);
 +      }
 +}
 ---
 
 
 >  >  --- pci/tsp_bus_mem.c  6 Feb 2012 02:14:15 -0000       1.12
 >  >  +++ pci/tsp_bus_mem.c  15 Jan 2014 15:18:22 -0000
  :
 >  >  -#define CHIP_MEM_SYS_START(v)    (((struct tsp_config *)(v))->pc_iobase)
 >  >  +#define CHIP_MEM_SYS_START(v)                                         \
 >  >  +      (((struct tsp_config *)(v))->pc_iobase | P_PCI_MEM)
  :
 >  I apply this patch.
 >  It work fine!
 
 Thanks. I'll commit this part soon.
 
 >  >return (alpha_btop(ALPHA_PHYS_TO_K0SEG(CHIP_MEM_SYS_START(v) + addr + 
 > off)));
 >  My patch is work but it is very very rare instances.
 >  I guess it should be MMU's magic!
 
 I guess some higher bits (around P_PCI_MEM) are ignored in pmap_enter()
 or alpha's MMU (but not sure).
 
 >  >  Can you any info why we need DEPTH_32 and
 >  
 >  Why I used to "32bpp display mode" because Xorg with wsfb driver is
 >  wrong on "8bpp display mode" and Xorg's radeon driver is wrong on alpha.
 >  I guess radeonfb_getcmap function and radeonfb_putcmap function hasn't
 >  been implemented yet.
 
 If this problem is not port specific, I wonder if we should make
 32BPP default..
 
 >  >  how does radeonfb_set_backlight() function fail?
 >  This patch is not necessary.
 
 Ok, I'll leave this part.
 
 ---
 Izumi Tsutsui