Merge branch 'for-jeff' of git://electric-eye.fr.zoreil.com/home/romieu/linux-2.6

[linux-2.6] / drivers / parisc / ccio-dma.c

/*
** ccio-dma.c:
**      DMA management routines for first generation cache-coherent machines.
**      Program U2/Uturn in "Virtual Mode" and use the I/O MMU.
**
**      (c) Copyright 2000 Grant Grundler
**      (c) Copyright 2000 Ryan Bradetich
**      (c) Copyright 2000 Hewlett-Packard Company
**
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
**
**
**  "Real Mode" operation refers to U2/Uturn chip operation.
**  U2/Uturn were designed to perform coherency checks w/o using
**  the I/O MMU - basically what x86 does.
**
**  Philipp Rumpf has a "Real Mode" driver for PCX-W machines at:
**      CVSROOT=:pserver:anonymous@198.186.203.37:/cvsroot/linux-parisc
**      cvs -z3 co linux/arch/parisc/kernel/dma-rm.c
**
**  I've rewritten his code to work under TPG's tree. See ccio-rm-dma.c.
**
**  Drawbacks of using Real Mode are:
**      o outbound DMA is slower - U2 won't prefetch data (GSC+ XQL signal).
**      o Inbound DMA less efficient - U2 can't use DMA_FAST attribute.
**      o Ability to do scatter/gather in HW is lost.
**      o Doesn't work under PCX-U/U+ machines since they didn't follow
**        the coherency design originally worked out. Only PCX-W does.
*/

#include <linux/config.h>
#include <linux/types.h>
#include <linux/init.h>
#include <linux/mm.h>
#include <linux/spinlock.h>
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/pci.h>
#include <linux/reboot.h>

#include <asm/byteorder.h>
#include <asm/cache.h>          /* for L1_CACHE_BYTES */
#include <asm/uaccess.h>
#include <asm/page.h>
#include <asm/dma.h>
#include <asm/io.h>
#include <asm/hardware.h>       /* for register_module() */
#include <asm/parisc-device.h>

/* 
** Choose "ccio" since that's what HP-UX calls it.
** Make it easier for folks to migrate from one to the other :^)
*/
#define MODULE_NAME "ccio"

#undef DEBUG_CCIO_RES
#undef DEBUG_CCIO_RUN
#undef DEBUG_CCIO_INIT
#undef DEBUG_CCIO_RUN_SG

#ifdef CONFIG_PROC_FS
/*
 * CCIO_SEARCH_TIME can help measure how fast the bitmap search is.
 * impacts performance though - ditch it if you don't use it.
 */
#define CCIO_SEARCH_TIME
#undef CCIO_MAP_STATS
#else
#undef CCIO_SEARCH_TIME
#undef CCIO_MAP_STATS
#endif

#include <linux/proc_fs.h>
#include <asm/runway.h>         /* for proc_runway_root */

#ifdef DEBUG_CCIO_INIT
#define DBG_INIT(x...)  printk(x)
#else
#define DBG_INIT(x...)
#endif

#ifdef DEBUG_CCIO_RUN
#define DBG_RUN(x...)   printk(x)
#else
#define DBG_RUN(x...)
#endif

#ifdef DEBUG_CCIO_RES
#define DBG_RES(x...)   printk(x)
#else
#define DBG_RES(x...)
#endif

#ifdef DEBUG_CCIO_RUN_SG
#define DBG_RUN_SG(x...) printk(x)
#else
#define DBG_RUN_SG(x...)
#endif

#define CCIO_INLINE     inline
#define WRITE_U32(value, addr) __raw_writel(value, addr)
#define READ_U32(addr) __raw_readl(addr)

#define U2_IOA_RUNWAY 0x580
#define U2_BC_GSC     0x501
#define UTURN_IOA_RUNWAY 0x581
#define UTURN_BC_GSC     0x502

#define IOA_NORMAL_MODE      0x00020080 /* IO_CONTROL to turn on CCIO        */
#define CMD_TLB_DIRECT_WRITE 35         /* IO_COMMAND for I/O TLB Writes     */
#define CMD_TLB_PURGE        33         /* IO_COMMAND to Purge I/O TLB entry */

struct ioa_registers {
        /* Runway Supervisory Set */
        int32_t    unused1[12];
        uint32_t   io_command;             /* Offset 12 */
        uint32_t   io_status;              /* Offset 13 */
        uint32_t   io_control;             /* Offset 14 */
        int32_t    unused2[1];

        /* Runway Auxiliary Register Set */
        uint32_t   io_err_resp;            /* Offset  0 */
        uint32_t   io_err_info;            /* Offset  1 */
        uint32_t   io_err_req;             /* Offset  2 */
        uint32_t   io_err_resp_hi;         /* Offset  3 */
        uint32_t   io_tlb_entry_m;         /* Offset  4 */
        uint32_t   io_tlb_entry_l;         /* Offset  5 */
        uint32_t   unused3[1];
        uint32_t   io_pdir_base;           /* Offset  7 */
        uint32_t   io_io_low_hv;           /* Offset  8 */
        uint32_t   io_io_high_hv;          /* Offset  9 */
        uint32_t   unused4[1];
        uint32_t   io_chain_id_mask;       /* Offset 11 */
        uint32_t   unused5[2];
        uint32_t   io_io_low;              /* Offset 14 */
        uint32_t   io_io_high;             /* Offset 15 */
};

/*
** IOA Registers
** -------------
**
** Runway IO_CONTROL Register (+0x38)
** 
** The Runway IO_CONTROL register controls the forwarding of transactions.
**
** | 0  ...  13  |  14 15 | 16 ... 21 | 22 | 23 24 |  25 ... 31 |
** |    HV       |   TLB  |  reserved | HV | mode  |  reserved  |
**
** o mode field indicates the address translation of transactions
**   forwarded from Runway to GSC+:
**       Mode Name     Value        Definition
**       Off (default)   0          Opaque to matching addresses.
**       Include         1          Transparent for matching addresses.
**       Peek            3          Map matching addresses.
**
**       + "Off" mode: Runway transactions which match the I/O range
**         specified by the IO_IO_LOW/IO_IO_HIGH registers will be ignored.
**       + "Include" mode: all addresses within the I/O range specified
**         by the IO_IO_LOW and IO_IO_HIGH registers are transparently
**         forwarded. This is the I/O Adapter's normal operating mode.
**       + "Peek" mode: used during system configuration to initialize the
**         GSC+ bus. Runway Write_Shorts in the address range specified by
**         IO_IO_LOW and IO_IO_HIGH are forwarded through the I/O Adapter
**         *AND* the GSC+ address is remapped to the Broadcast Physical
**         Address space by setting the 14 high order address bits of the
**         32 bit GSC+ address to ones.
**
** o TLB field affects transactions which are forwarded from GSC+ to Runway.
**   "Real" mode is the poweron default.
** 
**   TLB Mode  Value  Description
**   Real        0    No TLB translation. Address is directly mapped and the
**                    virtual address is composed of selected physical bits.
**   Error       1    Software fills the TLB manually.
**   Normal      2    IOA fetches IO TLB misses from IO PDIR (in host memory).
**
**
** IO_IO_LOW_HV   +0x60 (HV dependent)
** IO_IO_HIGH_HV  +0x64 (HV dependent)
** IO_IO_LOW      +0x78 (Architected register)
** IO_IO_HIGH     +0x7c (Architected register)
**
** IO_IO_LOW and IO_IO_HIGH set the lower and upper bounds of the
** I/O Adapter address space, respectively.
**
** 0  ... 7 | 8 ... 15 |  16   ...   31 |
** 11111111 | 11111111 |      address   |
**
** Each LOW/HIGH pair describes a disjoint address space region.
** (2 per GSC+ port). Each incoming Runway transaction address is compared
** with both sets of LOW/HIGH registers. If the address is in the range
** greater than or equal to IO_IO_LOW and less than IO_IO_HIGH the transaction
** for forwarded to the respective GSC+ bus.
** Specify IO_IO_LOW equal to or greater than IO_IO_HIGH to avoid specifying
** an address space region.
**
** In order for a Runway address to reside within GSC+ extended address space:
**      Runway Address [0:7]    must identically compare to 8'b11111111
**      Runway Address [8:11]   must be equal to IO_IO_LOW(_HV)[16:19]
**      Runway Address [12:23]  must be greater than or equal to
**                 IO_IO_LOW(_HV)[20:31] and less than IO_IO_HIGH(_HV)[20:31].
**      Runway Address [24:39]  is not used in the comparison.
**
** When the Runway transaction is forwarded to GSC+, the GSC+ address is
** as follows:
**      GSC+ Address[0:3]       4'b1111
**      GSC+ Address[4:29]      Runway Address[12:37]
**      GSC+ Address[30:31]     2'b00
**
** All 4 Low/High registers must be initialized (by PDC) once the lower bus
** is interrogated and address space is defined. The operating system will
** modify the architectural IO_IO_LOW and IO_IO_HIGH registers following
** the PDC initialization.  However, the hardware version dependent IO_IO_LOW
** and IO_IO_HIGH registers should not be subsequently altered by the OS.
** 
** Writes to both sets of registers will take effect immediately, bypassing
** the queues, which ensures that subsequent Runway transactions are checked
** against the updated bounds values. However reads are queued, introducing
** the possibility of a read being bypassed by a subsequent write to the same
** register. This sequence can be avoided by having software wait for read
** returns before issuing subsequent writes.
*/

struct ioc {
        struct ioa_registers __iomem *ioc_regs;  /* I/O MMU base address */
        u8  *res_map;                   /* resource map, bit == pdir entry */
        u64 *pdir_base;                 /* physical base address */
        u32 pdir_size;                  /* bytes, function of IOV Space size */
        u32 res_hint;                   /* next available IOVP - 
                                           circular search */
        u32 res_size;                   /* size of resource map in bytes */
        spinlock_t res_lock;

#ifdef CCIO_SEARCH_TIME
#define CCIO_SEARCH_SAMPLE 0x100
        unsigned long avg_search[CCIO_SEARCH_SAMPLE];
        unsigned long avg_idx;            /* current index into avg_search */
#endif
#ifdef CCIO_MAP_STATS
        unsigned long used_pages;
        unsigned long msingle_calls;
        unsigned long msingle_pages;
        unsigned long msg_calls;
        unsigned long msg_pages;
        unsigned long usingle_calls;
        unsigned long usingle_pages;
        unsigned long usg_calls;
        unsigned long usg_pages;
#endif
        unsigned short cujo20_bug;

        /* STUFF We don't need in performance path */
        u32 chainid_shift;              /* specify bit location of chain_id */
        struct ioc *next;               /* Linked list of discovered iocs */
        const char *name;               /* device name from firmware */
        unsigned int hw_path;           /* the hardware path this ioc is associatd with */
        struct pci_dev *fake_pci_dev;   /* the fake pci_dev for non-pci devs */
        struct resource mmio_region[2]; /* The "routed" MMIO regions */
};

static struct ioc *ioc_list;
static int ioc_count;

/**************************************************************
*
*   I/O Pdir Resource Management
*
*   Bits set in the resource map are in use.
*   Each bit can represent a number of pages.
*   LSbs represent lower addresses (IOVA's).
*
*   This was was copied from sba_iommu.c. Don't try to unify
*   the two resource managers unless a way to have different
*   allocation policies is also adjusted. We'd like to avoid
*   I/O TLB thrashing by having resource allocation policy
*   match the I/O TLB replacement policy.
*
***************************************************************/
#define IOVP_SIZE PAGE_SIZE
#define IOVP_SHIFT PAGE_SHIFT
#define IOVP_MASK PAGE_MASK

/* Convert from IOVP to IOVA and vice versa. */
#define CCIO_IOVA(iovp,offset) ((iovp) | (offset))
#define CCIO_IOVP(iova) ((iova) & IOVP_MASK)

#define PDIR_INDEX(iovp)    ((iovp)>>IOVP_SHIFT)
#define MKIOVP(pdir_idx)    ((long)(pdir_idx) << IOVP_SHIFT)
#define MKIOVA(iovp,offset) (dma_addr_t)((long)iovp | (long)offset)
#define ROUNDUP(x,y) ((x + ((y)-1)) & ~((y)-1))

/*
** Don't worry about the 150% average search length on a miss.
** If the search wraps around, and passes the res_hint, it will
** cause the kernel to panic anyhow.
*/
#define CCIO_SEARCH_LOOP(ioc, res_idx, mask, size)  \
       for(; res_ptr < res_end; ++res_ptr) { \
               if(0 == (*res_ptr & mask)) { \
                       *res_ptr |= mask; \
                       res_idx = (unsigned int)((unsigned long)res_ptr - (unsigned long)ioc->res_map); \
                       ioc->res_hint = res_idx + (size >> 3); \
                       goto resource_found; \
               } \
       }

#define CCIO_FIND_FREE_MAPPING(ioa, res_idx, mask, size) \
       u##size *res_ptr = (u##size *)&((ioc)->res_map[ioa->res_hint & ~((size >> 3) - 1)]); \
       u##size *res_end = (u##size *)&(ioc)->res_map[ioa->res_size]; \
       CCIO_SEARCH_LOOP(ioc, res_idx, mask, size); \
       res_ptr = (u##size *)&(ioc)->res_map[0]; \
       CCIO_SEARCH_LOOP(ioa, res_idx, mask, size);

/*
** Find available bit in this ioa's resource map.
** Use a "circular" search:
**   o Most IOVA's are "temporary" - avg search time should be small.
** o keep a history of what happened for debugging
** o KISS.
**
** Perf optimizations:
** o search for log2(size) bits at a time.
** o search for available resource bits using byte/word/whatever.
** o use different search for "large" (eg > 4 pages) or "very large"
**   (eg > 16 pages) mappings.
*/

/**
 * ccio_alloc_range - Allocate pages in the ioc's resource map.
 * @ioc: The I/O Controller.
 * @pages_needed: The requested number of pages to be mapped into the
 * I/O Pdir...
 *
 * This function searches the resource map of the ioc to locate a range
 * of available pages for the requested size.
 */
static int
ccio_alloc_range(struct ioc *ioc, size_t size)
{
        unsigned int pages_needed = size >> IOVP_SHIFT;
        unsigned int res_idx;
#ifdef CCIO_SEARCH_TIME
        unsigned long cr_start = mfctl(16);
#endif
        
        BUG_ON(pages_needed == 0);
        BUG_ON((pages_needed * IOVP_SIZE) > DMA_CHUNK_SIZE);
     
        DBG_RES("%s() size: %d pages_needed %d\n", 
                __FUNCTION__, size, pages_needed);

        /*
        ** "seek and ye shall find"...praying never hurts either...
        ** ggg sacrifices another 710 to the computer gods.
        */

        if (pages_needed <= 8) {
                /*
                 * LAN traffic will not thrash the TLB IFF the same NIC
                 * uses 8 adjacent pages to map seperate payload data.
                 * ie the same byte in the resource bit map.
                 */
#if 0
                /* FIXME: bit search should shift it's way through
                 * an unsigned long - not byte at a time. As it is now,
                 * we effectively allocate this byte to this mapping.
                 */
                unsigned long mask = ~(~0UL >> pages_needed);
                CCIO_FIND_FREE_MAPPING(ioc, res_idx, mask, 8);
#else
                CCIO_FIND_FREE_MAPPING(ioc, res_idx, 0xff, 8);
#endif
        } else if (pages_needed <= 16) {
                CCIO_FIND_FREE_MAPPING(ioc, res_idx, 0xffff, 16);
        } else if (pages_needed <= 32) {
                CCIO_FIND_FREE_MAPPING(ioc, res_idx, ~(unsigned int)0, 32);
#ifdef __LP64__
        } else if (pages_needed <= 64) {
                CCIO_FIND_FREE_MAPPING(ioc, res_idx, ~0UL, 64);
#endif
        } else {
                panic("%s: %s() Too many pages to map. pages_needed: %u\n",
                       __FILE__,  __FUNCTION__, pages_needed);
        }

        panic("%s: %s() I/O MMU is out of mapping resources.\n", __FILE__,
              __FUNCTION__);
        
resource_found:
        
        DBG_RES("%s() res_idx %d res_hint: %d\n",
                __FUNCTION__, res_idx, ioc->res_hint);

#ifdef CCIO_SEARCH_TIME
        {
                unsigned long cr_end = mfctl(16);
                unsigned long tmp = cr_end - cr_start;
                /* check for roll over */
                cr_start = (cr_end < cr_start) ?  -(tmp) : (tmp);
        }
        ioc->avg_search[ioc->avg_idx++] = cr_start;
        ioc->avg_idx &= CCIO_SEARCH_SAMPLE - 1;
#endif
#ifdef CCIO_MAP_STATS
        ioc->used_pages += pages_needed;
#endif
        /* 
        ** return the bit address.
        */
        return res_idx << 3;
}

#define CCIO_FREE_MAPPINGS(ioc, res_idx, mask, size) \
        u##size *res_ptr = (u##size *)&((ioc)->res_map[res_idx]); \
        BUG_ON((*res_ptr & mask) != mask); \
        *res_ptr &= ~(mask);

/**
 * ccio_free_range - Free pages from the ioc's resource map.
 * @ioc: The I/O Controller.
 * @iova: The I/O Virtual Address.
 * @pages_mapped: The requested number of pages to be freed from the
 * I/O Pdir.
 *
 * This function frees the resouces allocated for the iova.
 */
static void
ccio_free_range(struct ioc *ioc, dma_addr_t iova, unsigned long pages_mapped)
{
        unsigned long iovp = CCIO_IOVP(iova);
        unsigned int res_idx = PDIR_INDEX(iovp) >> 3;

        BUG_ON(pages_mapped == 0);
        BUG_ON((pages_mapped * IOVP_SIZE) > DMA_CHUNK_SIZE);
        BUG_ON(pages_mapped > BITS_PER_LONG);

        DBG_RES("%s():  res_idx: %d pages_mapped %d\n", 
                __FUNCTION__, res_idx, pages_mapped);

#ifdef CCIO_MAP_STATS
        ioc->used_pages -= pages_mapped;
#endif

        if(pages_mapped <= 8) {
#if 0
                /* see matching comments in alloc_range */
                unsigned long mask = ~(~0UL >> pages_mapped);
                CCIO_FREE_MAPPINGS(ioc, res_idx, mask, 8);
#else
                CCIO_FREE_MAPPINGS(ioc, res_idx, 0xff, 8);
#endif
        } else if(pages_mapped <= 16) {
                CCIO_FREE_MAPPINGS(ioc, res_idx, 0xffff, 16);
        } else if(pages_mapped <= 32) {
                CCIO_FREE_MAPPINGS(ioc, res_idx, ~(unsigned int)0, 32);
#ifdef __LP64__
        } else if(pages_mapped <= 64) {
                CCIO_FREE_MAPPINGS(ioc, res_idx, ~0UL, 64);
#endif
        } else {
                panic("%s:%s() Too many pages to unmap.\n", __FILE__,
                      __FUNCTION__);
        }
}

/****************************************************************
**
**          CCIO dma_ops support routines
**
*****************************************************************/

typedef unsigned long space_t;
#define KERNEL_SPACE 0

/*
** DMA "Page Type" and Hints 
** o if SAFE_DMA isn't set, mapping is for FAST_DMA. SAFE_DMA should be
**   set for subcacheline DMA transfers since we don't want to damage the
**   other part of a cacheline.
** o SAFE_DMA must be set for "memory" allocated via pci_alloc_consistent().
**   This bit tells U2 to do R/M/W for partial cachelines. "Streaming"
**   data can avoid this if the mapping covers full cache lines.
** o STOP_MOST is needed for atomicity across cachelines.
**   Apperently only "some EISA devices" need this.
**   Using CONFIG_ISA is hack. Only the IOA with EISA under it needs
**   to use this hint iff the EISA devices needs this feature.
**   According to the U2 ERS, STOP_MOST enabled pages hurt performance.
** o PREFETCH should *not* be set for cases like Multiple PCI devices
**   behind GSCtoPCI (dino) bus converter. Only one cacheline per GSC
**   device can be fetched and multiply DMA streams will thrash the
**   prefetch buffer and burn memory bandwidth. See 6.7.3 "Prefetch Rules
**   and Invalidation of Prefetch Entries".
**
** FIXME: the default hints need to be per GSC device - not global.
** 
** HP-UX dorks: linux device driver programming model is totally different
**    than HP-UX's. HP-UX always sets HINT_PREFETCH since it's drivers
**    do special things to work on non-coherent platforms...linux has to
**    be much more careful with this.
*/
#define IOPDIR_VALID    0x01UL
#define HINT_SAFE_DMA   0x02UL  /* used for pci_alloc_consistent() pages */
#ifdef CONFIG_EISA
#define HINT_STOP_MOST  0x04UL  /* LSL support */
#else
#define HINT_STOP_MOST  0x00UL  /* only needed for "some EISA devices" */
#endif
#define HINT_UDPATE_ENB 0x08UL  /* not used/supported by U2 */
#define HINT_PREFETCH   0x10UL  /* for outbound pages which are not SAFE */


/*
** Use direction (ie PCI_DMA_TODEVICE) to pick hint.
** ccio_alloc_consistent() depends on this to get SAFE_DMA
** when it passes in BIDIRECTIONAL flag.
*/
static u32 hint_lookup[] = {
        [PCI_DMA_BIDIRECTIONAL] = HINT_STOP_MOST | HINT_SAFE_DMA | IOPDIR_VALID,
        [PCI_DMA_TODEVICE]      = HINT_STOP_MOST | HINT_PREFETCH | IOPDIR_VALID,
        [PCI_DMA_FROMDEVICE]    = HINT_STOP_MOST | IOPDIR_VALID,
};

/**
 * ccio_io_pdir_entry - Initialize an I/O Pdir.
 * @pdir_ptr: A pointer into I/O Pdir.
 * @sid: The Space Identifier.
 * @vba: The virtual address.
 * @hints: The DMA Hint.
 *
 * Given a virtual address (vba, arg2) and space id, (sid, arg1),
 * load the I/O PDIR entry pointed to by pdir_ptr (arg0). Each IO Pdir
 * entry consists of 8 bytes as shown below (MSB == bit 0):
 *
 *
 * WORD 0:
 * +------+----------------+-----------------------------------------------+
 * | Phys | Virtual Index  |               Phys                            |
 * | 0:3  |     0:11       |               4:19                            |
 * |4 bits|   12 bits      |              16 bits                          |
 * +------+----------------+-----------------------------------------------+
 * WORD 1:
 * +-----------------------+-----------------------------------------------+
 * |      Phys    |  Rsvd  | Prefetch |Update |Rsvd  |Lock  |Safe  |Valid  |
 * |     20:39    |        | Enable   |Enable |      |Enable|DMA   |       |
 * |    20 bits   | 5 bits | 1 bit    |1 bit  |2 bits|1 bit |1 bit |1 bit  |
 * +-----------------------+-----------------------------------------------+
 *
 * The virtual index field is filled with the results of the LCI
 * (Load Coherence Index) instruction.  The 8 bits used for the virtual
 * index are bits 12:19 of the value returned by LCI.
 */ 
void CCIO_INLINE
ccio_io_pdir_entry(u64 *pdir_ptr, space_t sid, unsigned long vba,
                   unsigned long hints)
{
        register unsigned long pa;
        register unsigned long ci; /* coherent index */

        /* We currently only support kernel addresses */
        BUG_ON(sid != KERNEL_SPACE);

        mtsp(sid,1);

        /*
        ** WORD 1 - low order word
        ** "hints" parm includes the VALID bit!
        ** "dep" clobbers the physical address offset bits as well.
        */
        pa = virt_to_phys(vba);
        asm volatile("depw  %1,31,12,%0" : "+r" (pa) : "r" (hints));
        ((u32 *)pdir_ptr)[1] = (u32) pa;

        /*
        ** WORD 0 - high order word
        */

#ifdef __LP64__
        /*
        ** get bits 12:15 of physical address
        ** shift bits 16:31 of physical address
        ** and deposit them
        */
        asm volatile ("extrd,u %1,15,4,%0" : "=r" (ci) : "r" (pa));
        asm volatile ("extrd,u %1,31,16,%0" : "+r" (pa) : "r" (pa));
        asm volatile ("depd  %1,35,4,%0" : "+r" (pa) : "r" (ci));
#else
        pa = 0;
#endif
        /*
        ** get CPU coherency index bits
        ** Grab virtual index [0:11]
        ** Deposit virt_idx bits into I/O PDIR word
        */
        asm volatile ("lci %%r0(%%sr1, %1), %0" : "=r" (ci) : "r" (vba));
        asm volatile ("extru %1,19,12,%0" : "+r" (ci) : "r" (ci));
        asm volatile ("depw  %1,15,12,%0" : "+r" (pa) : "r" (ci));

        ((u32 *)pdir_ptr)[0] = (u32) pa;


        /* FIXME: PCX_W platforms don't need FDC/SYNC. (eg C360)
        **        PCX-U/U+ do. (eg C200/C240)
        **        PCX-T'? Don't know. (eg C110 or similar K-class)
        **
        ** See PDC_MODEL/option 0/SW_CAP word for "Non-coherent IO-PDIR bit".
        ** Hopefully we can patch (NOP) these out at boot time somehow.
        **
        ** "Since PCX-U employs an offset hash that is incompatible with
        ** the real mode coherence index generation of U2, the PDIR entry
        ** must be flushed to memory to retain coherence."
        */
        asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr));
        asm volatile("sync");
}

/**
 * ccio_clear_io_tlb - Remove stale entries from the I/O TLB.
 * @ioc: The I/O Controller.
 * @iovp: The I/O Virtual Page.
 * @byte_cnt: The requested number of bytes to be freed from the I/O Pdir.
 *
 * Purge invalid I/O PDIR entries from the I/O TLB.
 *
 * FIXME: Can we change the byte_cnt to pages_mapped?
 */
static CCIO_INLINE void
ccio_clear_io_tlb(struct ioc *ioc, dma_addr_t iovp, size_t byte_cnt)
{
        u32 chain_size = 1 << ioc->chainid_shift;

        iovp &= IOVP_MASK;      /* clear offset bits, just want pagenum */
        byte_cnt += chain_size;

        while(byte_cnt > chain_size) {
                WRITE_U32(CMD_TLB_PURGE | iovp, &ioc->ioc_regs->io_command);
                iovp += chain_size;
                byte_cnt -= chain_size;
        }
}

/**
 * ccio_mark_invalid - Mark the I/O Pdir entries invalid.
 * @ioc: The I/O Controller.
 * @iova: The I/O Virtual Address.
 * @byte_cnt: The requested number of bytes to be freed from the I/O Pdir.
 *
 * Mark the I/O Pdir entries invalid and blow away the corresponding I/O
 * TLB entries.
 *
 * FIXME: at some threshhold it might be "cheaper" to just blow
 *        away the entire I/O TLB instead of individual entries.
 *
 * FIXME: Uturn has 256 TLB entries. We don't need to purge every
 *        PDIR entry - just once for each possible TLB entry.
 *        (We do need to maker I/O PDIR entries invalid regardless).
 *
 * FIXME: Can we change byte_cnt to pages_mapped?
 */ 
static CCIO_INLINE void
ccio_mark_invalid(struct ioc *ioc, dma_addr_t iova, size_t byte_cnt)
{
        u32 iovp = (u32)CCIO_IOVP(iova);
        size_t saved_byte_cnt;

        /* round up to nearest page size */
        saved_byte_cnt = byte_cnt = ROUNDUP(byte_cnt, IOVP_SIZE);

        while(byte_cnt > 0) {
                /* invalidate one page at a time */
                unsigned int idx = PDIR_INDEX(iovp);
                char *pdir_ptr = (char *) &(ioc->pdir_base[idx]);

                BUG_ON(idx >= (ioc->pdir_size / sizeof(u64)));
                pdir_ptr[7] = 0;        /* clear only VALID bit */ 
                /*
                ** FIXME: PCX_W platforms don't need FDC/SYNC. (eg C360)
                **   PCX-U/U+ do. (eg C200/C240)
                ** See PDC_MODEL/option 0/SW_CAP for "Non-coherent IO-PDIR bit".
                **
                ** Hopefully someone figures out how to patch (NOP) the
                ** FDC/SYNC out at boot time.
                */
                asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr[7]));

                iovp     += IOVP_SIZE;
                byte_cnt -= IOVP_SIZE;
        }

        asm volatile("sync");
        ccio_clear_io_tlb(ioc, CCIO_IOVP(iova), saved_byte_cnt);
}

/****************************************************************
**
**          CCIO dma_ops
**
*****************************************************************/

/**
 * ccio_dma_supported - Verify the IOMMU supports the DMA address range.
 * @dev: The PCI device.
 * @mask: A bit mask describing the DMA address range of the device.
 *
 * This function implements the pci_dma_supported function.
 */
static int 
ccio_dma_supported(struct device *dev, u64 mask)
{
        if(dev == NULL) {
                printk(KERN_ERR MODULE_NAME ": EISA/ISA/et al not supported\n");
                BUG();
                return 0;
        }

        /* only support 32-bit devices (ie PCI/GSC) */
        return (int)(mask == 0xffffffffUL);
}

/**
 * ccio_map_single - Map an address range into the IOMMU.
 * @dev: The PCI device.
 * @addr: The start address of the DMA region.
 * @size: The length of the DMA region.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_map_single function.
 */
static dma_addr_t 
ccio_map_single(struct device *dev, void *addr, size_t size,
                enum dma_data_direction direction)
{
        int idx;
        struct ioc *ioc;
        unsigned long flags;
        dma_addr_t iovp;
        dma_addr_t offset;
        u64 *pdir_start;
        unsigned long hint = hint_lookup[(int)direction];

        BUG_ON(!dev);
        ioc = GET_IOC(dev);

        BUG_ON(size <= 0);

        /* save offset bits */
        offset = ((unsigned long) addr) & ~IOVP_MASK;

        /* round up to nearest IOVP_SIZE */
        size = ROUNDUP(size + offset, IOVP_SIZE);
        spin_lock_irqsave(&ioc->res_lock, flags);

#ifdef CCIO_MAP_STATS
        ioc->msingle_calls++;
        ioc->msingle_pages += size >> IOVP_SHIFT;
#endif

        idx = ccio_alloc_range(ioc, size);
        iovp = (dma_addr_t)MKIOVP(idx);

        pdir_start = &(ioc->pdir_base[idx]);

        DBG_RUN("%s() 0x%p -> 0x%lx size: %0x%x\n",
                __FUNCTION__, addr, (long)iovp | offset, size);

        /* If not cacheline aligned, force SAFE_DMA on the whole mess */
        if((size % L1_CACHE_BYTES) || ((unsigned long)addr % L1_CACHE_BYTES))
                hint |= HINT_SAFE_DMA;

        while(size > 0) {
                ccio_io_pdir_entry(pdir_start, KERNEL_SPACE, (unsigned long)addr, hint);

                DBG_RUN(" pdir %p %08x%08x\n",
                        pdir_start,
                        (u32) (((u32 *) pdir_start)[0]),
                        (u32) (((u32 *) pdir_start)[1]));
                ++pdir_start;
                addr += IOVP_SIZE;
                size -= IOVP_SIZE;
        }

        spin_unlock_irqrestore(&ioc->res_lock, flags);

        /* form complete address */
        return CCIO_IOVA(iovp, offset);
}

/**
 * ccio_unmap_single - Unmap an address range from the IOMMU.
 * @dev: The PCI device.
 * @addr: The start address of the DMA region.
 * @size: The length of the DMA region.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_unmap_single function.
 */
static void 
ccio_unmap_single(struct device *dev, dma_addr_t iova, size_t size, 
                  enum dma_data_direction direction)
{
        struct ioc *ioc;
        unsigned long flags; 
        dma_addr_t offset = iova & ~IOVP_MASK;
        
        BUG_ON(!dev);
        ioc = GET_IOC(dev);

        DBG_RUN("%s() iovp 0x%lx/%x\n",
                __FUNCTION__, (long)iova, size);

        iova ^= offset;        /* clear offset bits */
        size += offset;
        size = ROUNDUP(size, IOVP_SIZE);

        spin_lock_irqsave(&ioc->res_lock, flags);

#ifdef CCIO_MAP_STATS
        ioc->usingle_calls++;
        ioc->usingle_pages += size >> IOVP_SHIFT;
#endif

        ccio_mark_invalid(ioc, iova, size);
        ccio_free_range(ioc, iova, (size >> IOVP_SHIFT));
        spin_unlock_irqrestore(&ioc->res_lock, flags);
}

/**
 * ccio_alloc_consistent - Allocate a consistent DMA mapping.
 * @dev: The PCI device.
 * @size: The length of the DMA region.
 * @dma_handle: The DMA address handed back to the device (not the cpu).
 *
 * This function implements the pci_alloc_consistent function.
 */
static void * 
ccio_alloc_consistent(struct device *dev, size_t size, dma_addr_t *dma_handle, gfp_t flag)
{
      void *ret;
#if 0
/* GRANT Need to establish hierarchy for non-PCI devs as well
** and then provide matching gsc_map_xxx() functions for them as well.
*/
        if(!hwdev) {
                /* only support PCI */
                *dma_handle = 0;
                return 0;
        }
#endif
        ret = (void *) __get_free_pages(flag, get_order(size));

        if (ret) {
                memset(ret, 0, size);
                *dma_handle = ccio_map_single(dev, ret, size, PCI_DMA_BIDIRECTIONAL);
        }

        return ret;
}

/**
 * ccio_free_consistent - Free a consistent DMA mapping.
 * @dev: The PCI device.
 * @size: The length of the DMA region.
 * @cpu_addr: The cpu address returned from the ccio_alloc_consistent.
 * @dma_handle: The device address returned from the ccio_alloc_consistent.
 *
 * This function implements the pci_free_consistent function.
 */
static void 
ccio_free_consistent(struct device *dev, size_t size, void *cpu_addr, 
                     dma_addr_t dma_handle)
{
        ccio_unmap_single(dev, dma_handle, size, 0);
        free_pages((unsigned long)cpu_addr, get_order(size));
}

/*
** Since 0 is a valid pdir_base index value, can't use that
** to determine if a value is valid or not. Use a flag to indicate
** the SG list entry contains a valid pdir index.
*/
#define PIDE_FLAG 0x80000000UL

#ifdef CCIO_MAP_STATS
#define IOMMU_MAP_STATS
#endif
#include "iommu-helpers.h"

/**
 * ccio_map_sg - Map the scatter/gather list into the IOMMU.
 * @dev: The PCI device.
 * @sglist: The scatter/gather list to be mapped in the IOMMU.
 * @nents: The number of entries in the scatter/gather list.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_map_sg function.
 */
static int
ccio_map_sg(struct device *dev, struct scatterlist *sglist, int nents, 
            enum dma_data_direction direction)
{
        struct ioc *ioc;
        int coalesced, filled = 0;
        unsigned long flags;
        unsigned long hint = hint_lookup[(int)direction];
        unsigned long prev_len = 0, current_len = 0;
        int i;
        
        BUG_ON(!dev);
        ioc = GET_IOC(dev);
        
        DBG_RUN_SG("%s() START %d entries\n", __FUNCTION__, nents);

        /* Fast path single entry scatterlists. */
        if (nents == 1) {
                sg_dma_address(sglist) = ccio_map_single(dev,
                                (void *)sg_virt_addr(sglist), sglist->length,
                                direction);
                sg_dma_len(sglist) = sglist->length;
                return 1;
        }

        for(i = 0; i < nents; i++)
                prev_len += sglist[i].length;
        
        spin_lock_irqsave(&ioc->res_lock, flags);

#ifdef CCIO_MAP_STATS
        ioc->msg_calls++;
#endif

        /*
        ** First coalesce the chunks and allocate I/O pdir space
        **
        ** If this is one DMA stream, we can properly map using the
        ** correct virtual address associated with each DMA page.
        ** w/o this association, we wouldn't have coherent DMA!
        ** Access to the virtual address is what forces a two pass algorithm.
        */
        coalesced = iommu_coalesce_chunks(ioc, sglist, nents, ccio_alloc_range);

        /*
        ** Program the I/O Pdir
        **
        ** map the virtual addresses to the I/O Pdir
        ** o dma_address will contain the pdir index
        ** o dma_len will contain the number of bytes to map 
        ** o page/offset contain the virtual address.
        */
        filled = iommu_fill_pdir(ioc, sglist, nents, hint, ccio_io_pdir_entry);

        spin_unlock_irqrestore(&ioc->res_lock, flags);

        BUG_ON(coalesced != filled);

        DBG_RUN_SG("%s() DONE %d mappings\n", __FUNCTION__, filled);

        for (i = 0; i < filled; i++)
                current_len += sg_dma_len(sglist + i);

        BUG_ON(current_len != prev_len);

        return filled;
}

/**
 * ccio_unmap_sg - Unmap the scatter/gather list from the IOMMU.
 * @dev: The PCI device.
 * @sglist: The scatter/gather list to be unmapped from the IOMMU.
 * @nents: The number of entries in the scatter/gather list.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_unmap_sg function.
 */
static void 
ccio_unmap_sg(struct device *dev, struct scatterlist *sglist, int nents, 
              enum dma_data_direction direction)
{
        struct ioc *ioc;

        BUG_ON(!dev);
        ioc = GET_IOC(dev);

        DBG_RUN_SG("%s() START %d entries,  %08lx,%x\n",
                __FUNCTION__, nents, sg_virt_addr(sglist), sglist->length);

#ifdef CCIO_MAP_STATS
        ioc->usg_calls++;
#endif

        while(sg_dma_len(sglist) && nents--) {

#ifdef CCIO_MAP_STATS
                ioc->usg_pages += sg_dma_len(sglist) >> PAGE_SHIFT;
#endif
                ccio_unmap_single(dev, sg_dma_address(sglist),
                                  sg_dma_len(sglist), direction);
                ++sglist;
        }

        DBG_RUN_SG("%s() DONE (nents %d)\n", __FUNCTION__, nents);
}

static struct hppa_dma_ops ccio_ops = {
        .dma_supported =        ccio_dma_supported,
        .alloc_consistent =     ccio_alloc_consistent,
        .alloc_noncoherent =    ccio_alloc_consistent,
        .free_consistent =      ccio_free_consistent,
        .map_single =           ccio_map_single,
        .unmap_single =         ccio_unmap_single,
        .map_sg =               ccio_map_sg,
        .unmap_sg =             ccio_unmap_sg,
        .dma_sync_single_for_cpu =      NULL,   /* NOP for U2/Uturn */
        .dma_sync_single_for_device =   NULL,   /* NOP for U2/Uturn */
        .dma_sync_sg_for_cpu =          NULL,   /* ditto */
        .dma_sync_sg_for_device =               NULL,   /* ditto */
};

#ifdef CONFIG_PROC_FS
static int proc_append(char *src, int len, char **dst, off_t *offset, int *max)
{
        if (len < *offset) {
                *offset -= len;
                return 0;
        }
        if (*offset > 0) {
                src += *offset;
                len -= *offset;
                *offset = 0;
        }
        if (len > *max) {
                len = *max;
        }
        memcpy(*dst, src, len);
        *dst += len;
        *max -= len;
        return (*max == 0);
}

static int ccio_proc_info(char *buf, char **start, off_t offset, int count,
                          int *eof, void *data)
{
        int max = count;
        char tmp[80]; /* width of an ANSI-standard terminal */
        struct ioc *ioc = ioc_list;

        while (ioc != NULL) {
                unsigned int total_pages = ioc->res_size << 3;
                unsigned long avg = 0, min, max;
                int j, len;

                len = sprintf(tmp, "%s\n", ioc->name);
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
                
                len = sprintf(tmp, "Cujo 2.0 bug    : %s\n",
                              (ioc->cujo20_bug ? "yes" : "no"));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
                
                len = sprintf(tmp, "IO PDIR size    : %d bytes (%d entries)\n",
                              total_pages * 8, total_pages);
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
#ifdef CCIO_MAP_STATS
                len = sprintf(tmp, "IO PDIR entries : %ld free  %ld used (%d%%)\n",
                              total_pages - ioc->used_pages, ioc->used_pages,
                              (int)(ioc->used_pages * 100 / total_pages));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
#endif
                len = sprintf(tmp, "Resource bitmap : %d bytes (%d pages)\n", 
                        ioc->res_size, total_pages);
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
#ifdef CCIO_SEARCH_TIME
                min = max = ioc->avg_search[0];
                for(j = 0; j < CCIO_SEARCH_SAMPLE; ++j) {
                        avg += ioc->avg_search[j];
                        if(ioc->avg_search[j] > max) 
                                max = ioc->avg_search[j];
                        if(ioc->avg_search[j] < min) 
                                min = ioc->avg_search[j];
                }
                avg /= CCIO_SEARCH_SAMPLE;
                len = sprintf(tmp, "  Bitmap search : %ld/%ld/%ld (min/avg/max CPU Cycles)\n",
                              min, avg, max);
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
#endif
#ifdef CCIO_MAP_STATS
                len = sprintf(tmp, "pci_map_single(): %8ld calls  %8ld pages (avg %d/1000)\n",
                              ioc->msingle_calls, ioc->msingle_pages,
                              (int)((ioc->msingle_pages * 1000)/ioc->msingle_calls));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
                

                /* KLUGE - unmap_sg calls unmap_single for each mapped page */
                min = ioc->usingle_calls - ioc->usg_calls;
                max = ioc->usingle_pages - ioc->usg_pages;
                len = sprintf(tmp, "pci_unmap_single: %8ld calls  %8ld pages (avg %d/1000)\n",
                              min, max, (int)((max * 1000)/min));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
 
                len = sprintf(tmp, "pci_map_sg()    : %8ld calls  %8ld pages (avg %d/1000)\n",
                              ioc->msg_calls, ioc->msg_pages,
                              (int)((ioc->msg_pages * 1000)/ioc->msg_calls));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
                len = sprintf(tmp, "pci_unmap_sg()  : %8ld calls  %8ld pages (avg %d/1000)\n\n\n",
                              ioc->usg_calls, ioc->usg_pages,
                              (int)((ioc->usg_pages * 1000)/ioc->usg_calls));
                if (proc_append(tmp, len, &buf, &offset, &count))
                        break;
#endif  /* CCIO_MAP_STATS */
                ioc = ioc->next;
        }

        if (count == 0) {
                *eof = 1;
        }
        return (max - count);
}

static int ccio_resource_map(char *buf, char **start, off_t offset, int len,
                             int *eof, void *data)
{
        struct ioc *ioc = ioc_list;

        buf[0] = '\0';
        while (ioc != NULL) {
                u32 *res_ptr = (u32 *)ioc->res_map;
                int j;

                for (j = 0; j < (ioc->res_size / sizeof(u32)); j++) {
                        if ((j & 7) == 0)
                                strcat(buf,"\n   ");
                        sprintf(buf, "%s %08x", buf, *res_ptr);
                        res_ptr++;
                }
                strcat(buf, "\n\n");
                ioc = ioc->next;
                break; /* XXX - remove me */
        }

        return strlen(buf);
}
#endif

/**
 * ccio_find_ioc - Find the ioc in the ioc_list
 * @hw_path: The hardware path of the ioc.
 *
 * This function searches the ioc_list for an ioc that matches
 * the provide hardware path.
 */
static struct ioc * ccio_find_ioc(int hw_path)
{
        int i;
        struct ioc *ioc;

        ioc = ioc_list;
        for (i = 0; i < ioc_count; i++) {
                if (ioc->hw_path == hw_path)
                        return ioc;

                ioc = ioc->next;
        }

        return NULL;
}

/**
 * ccio_get_iommu - Find the iommu which controls this device
 * @dev: The parisc device.
 *
 * This function searches through the registered IOMMU's and returns
 * the appropriate IOMMU for the device based on its hardware path.
 */
void * ccio_get_iommu(const struct parisc_device *dev)
{
        dev = find_pa_parent_type(dev, HPHW_IOA);
        if (!dev)
                return NULL;

        return ccio_find_ioc(dev->hw_path);
}

#define CUJO_20_STEP       0x10000000   /* inc upper nibble */

/* Cujo 2.0 has a bug which will silently corrupt data being transferred
 * to/from certain pages.  To avoid this happening, we mark these pages
 * as `used', and ensure that nothing will try to allocate from them.
 */
void ccio_cujo20_fixup(struct parisc_device *cujo, u32 iovp)
{
        unsigned int idx;
        struct parisc_device *dev = parisc_parent(cujo);
        struct ioc *ioc = ccio_get_iommu(dev);
        u8 *res_ptr;

        ioc->cujo20_bug = 1;
        res_ptr = ioc->res_map;
        idx = PDIR_INDEX(iovp) >> 3;

        while (idx < ioc->res_size) {
                res_ptr[idx] |= 0xff;
                idx += PDIR_INDEX(CUJO_20_STEP) >> 3;
        }
}

#if 0
/* GRANT -  is this needed for U2 or not? */

/*
** Get the size of the I/O TLB for this I/O MMU.
**
** If spa_shift is non-zero (ie probably U2),
** then calculate the I/O TLB size using spa_shift.
**
** Otherwise we are supposed to get the IODC entry point ENTRY TLB
** and execute it. However, both U2 and Uturn firmware supplies spa_shift.
** I think only Java (K/D/R-class too?) systems don't do this.
*/
static int
ccio_get_iotlb_size(struct parisc_device *dev)
{
        if (dev->spa_shift == 0) {
                panic("%s() : Can't determine I/O TLB size.\n", __FUNCTION__);
        }
        return (1 << dev->spa_shift);
}
#else

/* Uturn supports 256 TLB entries */
#define CCIO_CHAINID_SHIFT      8
#define CCIO_CHAINID_MASK       0xff
#endif /* 0 */

/* We *can't* support JAVA (T600). Venture there at your own risk. */
static struct parisc_device_id ccio_tbl[] = {
        { HPHW_IOA, HVERSION_REV_ANY_ID, U2_IOA_RUNWAY, 0xb }, /* U2 */
        { HPHW_IOA, HVERSION_REV_ANY_ID, UTURN_IOA_RUNWAY, 0xb }, /* UTurn */
        { 0, }
};

static int ccio_probe(struct parisc_device *dev);

static struct parisc_driver ccio_driver = {
        .name =         "ccio",
        .id_table =     ccio_tbl,
        .probe =        ccio_probe,
};

/**
 * ccio_ioc_init - Initalize the I/O Controller
 * @ioc: The I/O Controller.
 *
 * Initalize the I/O Controller which includes setting up the
 * I/O Page Directory, the resource map, and initalizing the
 * U2/Uturn chip into virtual mode.
 */
static void
ccio_ioc_init(struct ioc *ioc)
{
        int i;
        unsigned int iov_order;
        u32 iova_space_size;

        /*
        ** Determine IOVA Space size from memory size.
        **
        ** Ideally, PCI drivers would register the maximum number
        ** of DMA they can have outstanding for each device they
        ** own.  Next best thing would be to guess how much DMA
        ** can be outstanding based on PCI Class/sub-class. Both
        ** methods still require some "extra" to support PCI
        ** Hot-Plug/Removal of PCI cards. (aka PCI OLARD).
        */

        iova_space_size = (u32) (num_physpages / count_parisc_driver(&ccio_driver));

        /* limit IOVA space size to 1MB-1GB */

        if (iova_space_size < (1 << (20 - PAGE_SHIFT))) {
                iova_space_size =  1 << (20 - PAGE_SHIFT);
#ifdef __LP64__
        } else if (iova_space_size > (1 << (30 - PAGE_SHIFT))) {
                iova_space_size =  1 << (30 - PAGE_SHIFT);
#endif
        }

        /*
        ** iova space must be log2() in size.
        ** thus, pdir/res_map will also be log2().
        */

        /* We could use larger page sizes in order to *decrease* the number
        ** of mappings needed.  (ie 8k pages means 1/2 the mappings).
        **
        ** Note: Grant Grunder says "Using 8k I/O pages isn't trivial either
        **   since the pages must also be physically contiguous - typically
        **   this is the case under linux."
        */

        iov_order = get_order(iova_space_size << PAGE_SHIFT);

        /* iova_space_size is now bytes, not pages */
        iova_space_size = 1 << (iov_order + PAGE_SHIFT);

        ioc->pdir_size = (iova_space_size / IOVP_SIZE) * sizeof(u64);

        BUG_ON(ioc->pdir_size > 8 * 1024 * 1024);   /* max pdir size <= 8MB */

        /* Verify it's a power of two */
        BUG_ON((1 << get_order(ioc->pdir_size)) != (ioc->pdir_size >> PAGE_SHIFT));

        DBG_INIT("%s() hpa 0x%p mem %luMB IOV %dMB (%d bits)\n",
                        __FUNCTION__, ioc->ioc_regs,
                        (unsigned long) num_physpages >> (20 - PAGE_SHIFT),
                        iova_space_size>>20,
                        iov_order + PAGE_SHIFT);

        ioc->pdir_base = (u64 *)__get_free_pages(GFP_KERNEL, 
                                                 get_order(ioc->pdir_size));
        if(NULL == ioc->pdir_base) {
                panic("%s() could not allocate I/O Page Table\n", __FUNCTION__);
        }
        memset(ioc->pdir_base, 0, ioc->pdir_size);

        BUG_ON((((unsigned long)ioc->pdir_base) & PAGE_MASK) != (unsigned long)ioc->pdir_base);
        DBG_INIT(" base %p\n", ioc->pdir_base);

        /* resource map size dictated by pdir_size */
        ioc->res_size = (ioc->pdir_size / sizeof(u64)) >> 3;
        DBG_INIT("%s() res_size 0x%x\n", __FUNCTION__, ioc->res_size);
        
        ioc->res_map = (u8 *)__get_free_pages(GFP_KERNEL, 
                                              get_order(ioc->res_size));
        if(NULL == ioc->res_map) {
                panic("%s() could not allocate resource map\n", __FUNCTION__);
        }
        memset(ioc->res_map, 0, ioc->res_size);

        /* Initialize the res_hint to 16 */
        ioc->res_hint = 16;

        /* Initialize the spinlock */
        spin_lock_init(&ioc->res_lock);

        /*
        ** Chainid is the upper most bits of an IOVP used to determine
        ** which TLB entry an IOVP will use.
        */
        ioc->chainid_shift = get_order(iova_space_size) + PAGE_SHIFT - CCIO_CHAINID_SHIFT;
        DBG_INIT(" chainid_shift 0x%x\n", ioc->chainid_shift);

        /*
        ** Initialize IOA hardware
        */
        WRITE_U32(CCIO_CHAINID_MASK << ioc->chainid_shift, 
                  &ioc->ioc_regs->io_chain_id_mask);

        WRITE_U32(virt_to_phys(ioc->pdir_base), 
                  &ioc->ioc_regs->io_pdir_base);

        /*
        ** Go to "Virtual Mode"
        */
        WRITE_U32(IOA_NORMAL_MODE, &ioc->ioc_regs->io_control);

        /*
        ** Initialize all I/O TLB entries to 0 (Valid bit off).
        */
        WRITE_U32(0, &ioc->ioc_regs->io_tlb_entry_m);
        WRITE_U32(0, &ioc->ioc_regs->io_tlb_entry_l);

        for(i = 1 << CCIO_CHAINID_SHIFT; i ; i--) {
                WRITE_U32((CMD_TLB_DIRECT_WRITE | (i << ioc->chainid_shift)),
                          &ioc->ioc_regs->io_command);
        }
}

static void
ccio_init_resource(struct resource *res, char *name, void __iomem *ioaddr)
{
        int result;

        res->parent = NULL;
        res->flags = IORESOURCE_MEM;
        /*
         * bracing ((signed) ...) are required for 64bit kernel because
         * we only want to sign extend the lower 16 bits of the register.
         * The upper 16-bits of range registers are hardcoded to 0xffff.
         */
        res->start = (unsigned long)((signed) READ_U32(ioaddr) << 16);
        res->end = (unsigned long)((signed) (READ_U32(ioaddr + 4) << 16) - 1);
        res->name = name;
        /*
         * Check if this MMIO range is disable
         */
        if (res->end + 1 == res->start)
                return;

        /* On some platforms (e.g. K-Class), we have already registered
         * resources for devices reported by firmware. Some are children
         * of ccio.
         * "insert" ccio ranges in the mmio hierarchy (/proc/iomem).
         */
        result = insert_resource(&iomem_resource, res);
        if (result < 0) {
                printk(KERN_ERR "%s() failed to claim CCIO bus address space (%08lx,%08lx)\n", 
                        __FUNCTION__, res->start, res->end);
        }
}

static void __init ccio_init_resources(struct ioc *ioc)
{
        struct resource *res = ioc->mmio_region;
        char *name = kmalloc(14, GFP_KERNEL);

        snprintf(name, 14, "GSC Bus [%d/]", ioc->hw_path);

        ccio_init_resource(res, name, &ioc->ioc_regs->io_io_low);
        ccio_init_resource(res + 1, name, &ioc->ioc_regs->io_io_low_hv);
}

static int new_ioc_area(struct resource *res, unsigned long size,
                unsigned long min, unsigned long max, unsigned long align)
{
        if (max <= min)
                return -EBUSY;

        res->start = (max - size + 1) &~ (align - 1);
        res->end = res->start + size;
        
        /* We might be trying to expand the MMIO range to include
         * a child device that has already registered it's MMIO space.
         * Use "insert" instead of request_resource().
         */
        if (!insert_resource(&iomem_resource, res))
                return 0;

        return new_ioc_area(res, size, min, max - size, align);
}

static int expand_ioc_area(struct resource *res, unsigned long size,
                unsigned long min, unsigned long max, unsigned long align)
{
        unsigned long start, len;

        if (!res->parent)
                return new_ioc_area(res, size, min, max, align);

        start = (res->start - size) &~ (align - 1);
        len = res->end - start + 1;
        if (start >= min) {
                if (!adjust_resource(res, start, len))
                        return 0;
        }

        start = res->start;
        len = ((size + res->end + align) &~ (align - 1)) - start;
        if (start + len <= max) {
                if (!adjust_resource(res, start, len))
                        return 0;
        }

        return -EBUSY;
}

/*
 * Dino calls this function.  Beware that we may get called on systems
 * which have no IOC (725, B180, C160L, etc) but do have a Dino.
 * So it's legal to find no parent IOC.
 *
 * Some other issues: one of the resources in the ioc may be unassigned.
 */
int ccio_allocate_resource(const struct parisc_device *dev,
                struct resource *res, unsigned long size,
                unsigned long min, unsigned long max, unsigned long align)
{
        struct resource *parent = &iomem_resource;
        struct ioc *ioc = ccio_get_iommu(dev);
        if (!ioc)
                goto out;

        parent = ioc->mmio_region;
        if (parent->parent &&
            !allocate_resource(parent, res, size, min, max, align, NULL, NULL))
                return 0;

        if ((parent + 1)->parent &&
            !allocate_resource(parent + 1, res, size, min, max, align,
                                NULL, NULL))
                return 0;

        if (!expand_ioc_area(parent, size, min, max, align)) {
                __raw_writel(((parent->start)>>16) | 0xffff0000,
                             &ioc->ioc_regs->io_io_low);
                __raw_writel(((parent->end)>>16) | 0xffff0000,
                             &ioc->ioc_regs->io_io_high);
        } else if (!expand_ioc_area(parent + 1, size, min, max, align)) {
                parent++;
                __raw_writel(((parent->start)>>16) | 0xffff0000,
                             &ioc->ioc_regs->io_io_low_hv);
                __raw_writel(((parent->end)>>16) | 0xffff0000,
                             &ioc->ioc_regs->io_io_high_hv);
        } else {
                return -EBUSY;
        }

 out:
        return allocate_resource(parent, res, size, min, max, align, NULL,NULL);
}

int ccio_request_resource(const struct parisc_device *dev,
                struct resource *res)
{
        struct resource *parent;
        struct ioc *ioc = ccio_get_iommu(dev);

        if (!ioc) {
                parent = &iomem_resource;
        } else if ((ioc->mmio_region->start <= res->start) &&
                        (res->end <= ioc->mmio_region->end)) {
                parent = ioc->mmio_region;
        } else if (((ioc->mmio_region + 1)->start <= res->start) &&
                        (res->end <= (ioc->mmio_region + 1)->end)) {
                parent = ioc->mmio_region + 1;
        } else {
                return -EBUSY;
        }

        /* "transparent" bus bridges need to register MMIO resources
         * firmware assigned them. e.g. children of hppb.c (e.g. K-class)
         * registered their resources in the PDC "bus walk" (See
         * arch/parisc/kernel/inventory.c).
         */
        return insert_resource(parent, res);
}

/**
 * ccio_probe - Determine if ccio should claim this device.
 * @dev: The device which has been found
 *
 * Determine if ccio should claim this chip (return 0) or not (return 1).
 * If so, initialize the chip and tell other partners in crime they
 * have work to do.
 */
static int ccio_probe(struct parisc_device *dev)
{
        int i;
        struct ioc *ioc, **ioc_p = &ioc_list;
        
        ioc = kzalloc(sizeof(struct ioc), GFP_KERNEL);
        if (ioc == NULL) {
                printk(KERN_ERR MODULE_NAME ": memory allocation failure\n");
                return 1;
        }

        ioc->name = dev->id.hversion == U2_IOA_RUNWAY ? "U2" : "UTurn";

        printk(KERN_INFO "Found %s at 0x%lx\n", ioc->name, dev->hpa.start);

        for (i = 0; i < ioc_count; i++) {
                ioc_p = &(*ioc_p)->next;
        }
        *ioc_p = ioc;

        ioc->hw_path = dev->hw_path;
        ioc->ioc_regs = ioremap(dev->hpa.start, 4096);
        ccio_ioc_init(ioc);
        ccio_init_resources(ioc);
        hppa_dma_ops = &ccio_ops;
        dev->dev.platform_data = kzalloc(sizeof(struct pci_hba_data), GFP_KERNEL);

        /* if this fails, no I/O cards will work, so may as well bug */
        BUG_ON(dev->dev.platform_data == NULL);
        HBA_DATA(dev->dev.platform_data)->iommu = ioc;
        

        if (ioc_count == 0) {
                /* FIXME: Create separate entries for each ioc */
                create_proc_read_entry(MODULE_NAME, S_IRWXU, proc_runway_root,
                                       ccio_proc_info, NULL);
                create_proc_read_entry(MODULE_NAME"-bitmap", S_IRWXU,
                                       proc_runway_root, ccio_resource_map, NULL);
        }

        ioc_count++;

        parisc_vmerge_boundary = IOVP_SIZE;
        parisc_vmerge_max_size = BITS_PER_LONG * IOVP_SIZE;
        parisc_has_iommu();
        return 0;
}

/**
 * ccio_init - ccio initalization procedure.
 *
 * Register this driver.
 */
void __init ccio_init(void)
{
        register_parisc_driver(&ccio_driver);
}