1[ NOTE: The virt_to_bus() and bus_to_virt() functions have been
   2        superseded by the functionality provided by the PCI DMA interface
   3        (see Documentation/DMA-API-HOWTO.txt).  They continue
   4        to be documented below for historical purposes, but new code
   5        must not use them. --davidm 00/12/12 ]
   7[ This is a mail message in response to a query on IO mapping, thus the
   8  strange format for a "document" ]
  10The AHA-1542 is a bus-master device, and your patch makes the driver give the
  11controller the physical address of the buffers, which is correct on x86
  12(because all bus master devices see the physical memory mappings directly). 
  14However, on many setups, there are actually _three_ different ways of looking
  15at memory addresses, and in this case we actually want the third, the
  16so-called "bus address". 
  18Essentially, the three ways of addressing memory are (this is "real memory",
  19that is, normal RAM--see later about other details): 
  21 - CPU untranslated.  This is the "physical" address.  Physical address 
  22   0 is what the CPU sees when it drives zeroes on the memory bus.
  24 - CPU translated address. This is the "virtual" address, and is 
  25   completely internal to the CPU itself with the CPU doing the appropriate
  26   translations into "CPU untranslated". 
  28 - bus address. This is the address of memory as seen by OTHER devices, 
  29   not the CPU. Now, in theory there could be many different bus 
  30   addresses, with each device seeing memory in some device-specific way, but
  31   happily most hardware designers aren't actually actively trying to make
  32   things any more complex than necessary, so you can assume that all 
  33   external hardware sees the memory the same way. 
  35Now, on normal PCs the bus address is exactly the same as the physical
  36address, and things are very simple indeed. However, they are that simple
  37because the memory and the devices share the same address space, and that is
  38not generally necessarily true on other PCI/ISA setups. 
  40Now, just as an example, on the PReP (PowerPC Reference Platform), the 
  41CPU sees a memory map something like this (this is from memory):
  43        0-2 GB          "real memory"
  44        2 GB-3 GB       "system IO" (inb/out and similar accesses on x86)
  45        3 GB-4 GB       "IO memory" (shared memory over the IO bus)
  47Now, that looks simple enough. However, when you look at the same thing from
  48the viewpoint of the devices, you have the reverse, and the physical memory
  49address 0 actually shows up as address 2 GB for any IO master.
  51So when the CPU wants any bus master to write to physical memory 0, it 
  52has to give the master address 0x80000000 as the memory address.
  54So, for example, depending on how the kernel is actually mapped on the 
  55PPC, you can end up with a setup like this:
  57 physical address:      0
  58 virtual address:       0xC0000000
  59 bus address:           0x80000000
  61where all the addresses actually point to the same thing.  It's just seen 
  62through different translations..
  64Similarly, on the Alpha, the normal translation is
  66 physical address:      0
  67 virtual address:       0xfffffc0000000000
  68 bus address:           0x40000000
  70(but there are also Alphas where the physical address and the bus address
  71are the same). 
  73Anyway, the way to look up all these translations, you do
  75        #include <asm/io.h>
  77        phys_addr = virt_to_phys(virt_addr);
  78        virt_addr = phys_to_virt(phys_addr);
  79         bus_addr = virt_to_bus(virt_addr);
  80        virt_addr = bus_to_virt(bus_addr);
  82Now, when do you need these?
  84You want the _virtual_ address when you are actually going to access that 
  85pointer from the kernel. So you can have something like this:
  87        /*
  88         * this is the hardware "mailbox" we use to communicate with
  89         * the controller. The controller sees this directly.
  90         */
  91        struct mailbox {
  92                __u32 status;
  93                __u32 bufstart;
  94                __u32 buflen;
  95                ..
  96        } mbox;
  98                unsigned char * retbuffer;
 100                /* get the address from the controller */
 101                retbuffer = bus_to_virt(mbox.bufstart);
 102                switch (retbuffer[0]) {
 103                        case STATUS_OK:
 104                                ...
 106on the other hand, you want the bus address when you have a buffer that 
 107you want to give to the controller:
 109        /* ask the controller to read the sense status into "sense_buffer" */
 110        mbox.bufstart = virt_to_bus(&sense_buffer);
 111        mbox.buflen = sizeof(sense_buffer);
 112        mbox.status = 0;
 113        notify_controller(&mbox);
 115And you generally _never_ want to use the physical address, because you can't
 116use that from the CPU (the CPU only uses translated virtual addresses), and
 117you can't use it from the bus master. 
 119So why do we care about the physical address at all? We do need the physical
 120address in some cases, it's just not very often in normal code.  The physical
 121address is needed if you use memory mappings, for example, because the
 122"remap_pfn_range()" mm function wants the physical address of the memory to
 123be remapped as measured in units of pages, a.k.a. the pfn (the memory
 124management layer doesn't know about devices outside the CPU, so it
 125shouldn't need to know about "bus addresses" etc).
 127NOTE NOTE NOTE! The above is only one part of the whole equation. The above
 128only talks about "real memory", that is, CPU memory (RAM). 
 130There is a completely different type of memory too, and that's the "shared
 131memory" on the PCI or ISA bus. That's generally not RAM (although in the case
 132of a video graphics card it can be normal DRAM that is just used for a frame
 133buffer), but can be things like a packet buffer in a network card etc. 
 135This memory is called "PCI memory" or "shared memory" or "IO memory" or
 136whatever, and there is only one way to access it: the readb/writeb and
 137related functions. You should never take the address of such memory, because
 138there is really nothing you can do with such an address: it's not
 139conceptually in the same memory space as "real memory" at all, so you cannot
 140just dereference a pointer. (Sadly, on x86 it _is_ in the same memory space,
 141so on x86 it actually works to just deference a pointer, but it's not
 144For such memory, you can do things like
 146 - reading:
 147        /*
 148         * read first 32 bits from ISA memory at 0xC0000, aka
 149         * C000:0000 in DOS terms
 150         */
 151        unsigned int signature = isa_readl(0xC0000);
 153 - remapping and writing:
 154        /*
 155         * remap framebuffer PCI memory area at 0xFC000000,
 156         * size 1MB, so that we can access it: We can directly
 157         * access only the 640k-1MB area, so anything else
 158         * has to be remapped.
 159         */
 160        void __iomem *baseptr = ioremap(0xFC000000, 1024*1024);
 162        /* write a 'A' to the offset 10 of the area */
 163        writeb('A',baseptr+10);
 165        /* unmap when we unload the driver */
 166        iounmap(baseptr);
 168 - copying and clearing:
 169        /* get the 6-byte Ethernet address at ISA address E000:0040 */
 170        memcpy_fromio(kernel_buffer, 0xE0040, 6);
 171        /* write a packet to the driver */
 172        memcpy_toio(0xE1000, skb->data, skb->len);
 173        /* clear the frame buffer */
 174        memset_io(0xA0000, 0, 0x10000);
 176OK, that just about covers the basics of accessing IO portably.  Questions?
 177Comments? You may think that all the above is overly complex, but one day you
 178might find yourself with a 500 MHz Alpha in front of you, and then you'll be
 179happy that your driver works ;)
 181Note that kernel versions 2.0.x (and earlier) mistakenly called the
 182ioremap() function "vremap()".  ioremap() is the proper name, but I
 183didn't think straight when I wrote it originally.  People who have to
 184support both can do something like:
 186        /* support old naming silliness */
 187        #if LINUX_VERSION_CODE < 0x020100                                     
 188        #define ioremap vremap
 189        #define iounmap vfree                                                     
 190        #endif
 192at the top of their source files, and then they can use the right names
 193even on 2.0.x systems. 
 195And the above sounds worse than it really is.  Most real drivers really
 196don't do all that complex things (or rather: the complexity is not so
 197much in the actual IO accesses as in error handling and timeouts etc). 
 198It's generally not hard to fix drivers, and in many cases the code
 199actually looks better afterwards:
 201        unsigned long signature = *(unsigned int *) 0xC0000;
 202                vs
 203        unsigned long signature = readl(0xC0000);
 205I think the second version actually is more readable, no?
 207                Linus