Date: Thu, 25 Sep 1997 20:54:05 +0200 (MET DST) From: Wout Klaren To: linux-m68k@phil.uni-sb.de Subject: Re: L68K: More patches for the Hades In-Reply-To: <9709250843.AA19172@faui21.informatik.uni-erlangen.de> Sender: owner-linux-m68k@phil.uni-sb.de Reply-To: Wout Klaren On Thu, 25 Sep 1997, Roman Hodek wrote: > Some comments to your patches: > > > +Hades support > > +CONFIG_HADES > > + This option enables support for the Hades Atari clone. If you say > > + Y here, PCI bus support will be compiled in and the TT SCSI DMA > > + emulation option for SCSI will be enabled. Note that a kernel > > + without this option set (but with Atari support) will also work > > + on the Hades, but without PCI bus support and SCSI transfers will > > + be very slow. If you plan to use this kernel on a Hades, say Y > > + here; otherwise say N. > > + > > This is not really true... The CONFIG_PCI and CONFIG_TT_DMA_EMUL are > extra config vars, that are only offered if CONFIG_HADES is selected, > too. > > CONFIG_HADES also does a few other things: It takes care that the > autolevel 2 interrupt isn't disabled, and the new virt_to_bus mapping. Ok, I have updated the comment in the patches below. > Do you need the VME mapping on a Hades? (AFAICR, it doesn't have a VME > bus.) Then you could use the tt1 reg for your mapping instead of tt0 > and thus save one of the transp. translation register sets. Just put > some "if (is_hades)" into the code where the VME mapping is set up and > use a different tt value then. The Hades has a VME bus, so tt1 is not free. The first versions of the PCI driver did not use tt0, but used kernel_map to remap the PCI memory, I/O and configuration areas. But it is easier to have the PCI bus at a fixed address. > To the PCI stuff: Do you know Torsten Lang? He has written some kind > of Hades PCI BIOS for mw. Unfortunately he can't give away the source > code (it's proprietary), but he surely can help you with PCI problems > in all ways... Yes, I know him. I have spoken with him about the PCI driver. > I'm not sure about these... can't they be used also somewhere else > where byteswapping wouldn't be good? Because of the other reactions in this mailing list, I have deleted them. When I added them I looked at the Linux/PPC port. In this port these macro's swap the byteorder. The Linux/PPC port has also versions which do not change the byteorder, but these macro's have other names. > You should better also add a runtime test for Hades, otherwise kernels > compiled with CONFIG_HADES don't run on other machines anymore... I.e., > > #define virt_to_bus(a) (virt_to_phys(a) + (is_hades ? 0x80000000 : 0)) You are right. I have fixed it in the pacthes below. Wout Klaren diff -ruN -x *.o -x *.a -x .*depend -x .config* -x .menuconfig* -x System.map -x .version -x autoconf.h -x compile.h -x vmlinux -x *_MODULES -x *.ver linux-2.1.55.orig/Documentation/Configure.help linux-2.1.55/Documentation/Configure.help --- linux-2.1.55.orig/Documentation/Configure.help Wed Sep 24 23:22:19 1997 +++ linux-2.1.55/Documentation/Configure.help Thu Sep 25 19:51:45 1997 @@ -5602,13 +5602,8 @@ Hades support CONFIG_HADES - This option enables support for the Hades Atari clone. If you say - Y here, PCI bus support will be compiled in and the TT SCSI DMA - emulation option for SCSI will be enabled. Note that a kernel - without this option set (but with Atari support) will also work - on the Hades, but without PCI bus support and SCSI transfers will - be very slow. If you plan to use this kernel on a Hades, say Y - here; otherwise say N. + This option enables support for the Hades Atari clone. If you plan to + use this kernel on a Hades, say Y here; otherwise say N. Macintosh support CONFIG_MAC diff -ruN -x *.o -x *.a -x .*depend -x .config* -x .menuconfig* -x System.map -x .version -x autoconf.h -x compile.h -x vmlinux -x *_MODULES -x *.ver linux-2.1.55.orig/include/asm-m68k/io.h linux-2.1.55/include/asm-m68k/io.h --- linux-2.1.55.orig/include/asm-m68k/io.h Wed Sep 24 23:22:21 1997 +++ linux-2.1.55/include/asm-m68k/io.h Thu Sep 25 19:39:57 1997 @@ -41,28 +41,6 @@ #define outb(x,addr) ((void) writeb(x,addr)) #define outb_p(x,addr) outb(x,addr) -extern inline unsigned short inw(int port) -{ - return le16_to_cpu(readw(port)); -} - -extern inline unsigned long inl(int port) -{ - return le32_to_cpu(readl(port)); -} - -extern inline unsigned short outw(unsigned short val, int port) -{ - writew(cpu_to_le16(val), port); - return val; -} - -extern inline unsigned long outl(unsigned long val, int port) -{ - writel(cpu_to_le32(val), port); - return val; -} - /* * Change virtual addresses to physical addresses and vv. */ @@ -84,8 +62,8 @@ * except on the PCI bus of the Hades. */ #ifdef CONFIG_HADES -#define virt_to_bus(a) (virt_to_phys(a) + 0x80000000) -#define bus_to_virt(a) (phys_to_virt((a) - 0x80000000)) +#define virt_to_bus(a) (virt_to_phys(a) + (is_hades ? 0x80000000 : 0)) +#define bus_to_virt(a) (phys_to_virt((a) - (is_hades ? 0x80000000 : 0))) #else #define virt_to_bus virt_to_phys #define bus_to_virt phys_to_virt