diff options
Diffstat (limited to 'drivers/parisc/ccio-dma.c')
| -rw-r--r-- | drivers/parisc/ccio-dma.c | 460 |
1 files changed, 231 insertions, 229 deletions
diff --git a/drivers/parisc/ccio-dma.c b/drivers/parisc/ccio-dma.c index 0e98a9d9834..8b490d77054 100644 --- a/drivers/parisc/ccio-dma.c +++ b/drivers/parisc/ccio-dma.c @@ -31,8 +31,8 @@ ** the coherency design originally worked out. Only PCX-W does. */ -#include <linux/config.h> #include <linux/types.h> +#include <linux/kernel.h> #include <linux/init.h> #include <linux/mm.h> #include <linux/spinlock.h> @@ -40,6 +40,11 @@ #include <linux/string.h> #include <linux/pci.h> #include <linux/reboot.h> +#include <linux/proc_fs.h> +#include <linux/seq_file.h> +#include <linux/scatterlist.h> +#include <linux/iommu-helper.h> +#include <linux/export.h> #include <asm/byteorder.h> #include <asm/cache.h> /* for L1_CACHE_BYTES */ @@ -62,18 +67,10 @@ #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 +/* depends on proc fs support. But costs CPU performance. */ +#undef CCIO_COLLECT_STATS #endif -#include <linux/proc_fs.h> #include <asm/runway.h> /* for proc_runway_root */ #ifdef DEBUG_CCIO_INIT @@ -100,9 +97,9 @@ #define DBG_RUN_SG(x...) #endif -#define CCIO_INLINE /* inline */ -#define WRITE_U32(value, addr) gsc_writel(value, (u32 *)(addr)) -#define READ_U32(addr) gsc_readl((u32 *)(addr)) +#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 @@ -115,28 +112,28 @@ struct ioa_registers { /* Runway Supervisory Set */ - volatile int32_t unused1[12]; - volatile uint32_t io_command; /* Offset 12 */ - volatile uint32_t io_status; /* Offset 13 */ - volatile uint32_t io_control; /* Offset 14 */ - volatile int32_t unused2[1]; + 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 */ - volatile uint32_t io_err_resp; /* Offset 0 */ - volatile uint32_t io_err_info; /* Offset 1 */ - volatile uint32_t io_err_req; /* Offset 2 */ - volatile uint32_t io_err_resp_hi; /* Offset 3 */ - volatile uint32_t io_tlb_entry_m; /* Offset 4 */ - volatile uint32_t io_tlb_entry_l; /* Offset 5 */ - volatile uint32_t unused3[1]; - volatile uint32_t io_pdir_base; /* Offset 7 */ - volatile uint32_t io_io_low_hv; /* Offset 8 */ - volatile uint32_t io_io_high_hv; /* Offset 9 */ - volatile uint32_t unused4[1]; - volatile uint32_t io_chain_id_mask; /* Offset 11 */ - volatile uint32_t unused5[2]; - volatile uint32_t io_io_low; /* Offset 14 */ - volatile uint32_t io_io_high; /* Offset 15 */ + 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 */ }; /* @@ -226,7 +223,7 @@ struct ioa_registers { */ struct ioc { - struct ioa_registers *ioc_hpa; /* I/O MMU base address */ + 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 */ @@ -235,12 +232,10 @@ struct ioc { u32 res_size; /* size of resource map in bytes */ spinlock_t res_lock; -#ifdef CCIO_SEARCH_TIME +#ifdef CCIO_COLLECT_STATS #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; @@ -291,7 +286,6 @@ static int ioc_count; #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. @@ -300,13 +294,17 @@ static int ioc_count; */ #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; \ - } \ - } + int ret;\ + unsigned int idx;\ + idx = (unsigned int)((unsigned long)res_ptr - (unsigned long)ioc->res_map); \ + ret = iommu_is_span_boundary(idx << 3, pages_needed, 0, boundary_size);\ + if ((0 == (*res_ptr & mask)) && !ret) { \ + *res_ptr |= mask; \ + res_idx = idx;\ + 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)]); \ @@ -339,11 +337,12 @@ static int ioc_count; * of available pages for the requested size. */ static int -ccio_alloc_range(struct ioc *ioc, size_t size) +ccio_alloc_range(struct ioc *ioc, struct device *dev, size_t size) { unsigned int pages_needed = size >> IOVP_SHIFT; unsigned int res_idx; -#ifdef CCIO_SEARCH_TIME + unsigned long boundary_size; +#ifdef CCIO_COLLECT_STATS unsigned long cr_start = mfctl(16); #endif @@ -351,17 +350,20 @@ ccio_alloc_range(struct ioc *ioc, size_t size) BUG_ON((pages_needed * IOVP_SIZE) > DMA_CHUNK_SIZE); DBG_RES("%s() size: %d pages_needed %d\n", - __FUNCTION__, size, pages_needed); + __func__, size, pages_needed); /* ** "seek and ye shall find"...praying never hurts either... ** ggg sacrifices another 710 to the computer gods. */ + boundary_size = ALIGN((unsigned long long)dma_get_seg_boundary(dev) + 1, + 1ULL << IOVP_SHIFT) >> IOVP_SHIFT; + if (pages_needed <= 8) { /* * LAN traffic will not thrash the TLB IFF the same NIC - * uses 8 adjacent pages to map seperate payload data. + * uses 8 adjacent pages to map separate payload data. * ie the same byte in the resource bit map. */ #if 0 @@ -384,18 +386,18 @@ ccio_alloc_range(struct ioc *ioc, size_t size) #endif } else { panic("%s: %s() Too many pages to map. pages_needed: %u\n", - __FILE__, __FUNCTION__, pages_needed); + __FILE__, __func__, pages_needed); } panic("%s: %s() I/O MMU is out of mapping resources.\n", __FILE__, - __FUNCTION__); + __func__); resource_found: DBG_RES("%s() res_idx %d res_hint: %d\n", - __FUNCTION__, res_idx, ioc->res_hint); + __func__, res_idx, ioc->res_hint); -#ifdef CCIO_SEARCH_TIME +#ifdef CCIO_COLLECT_STATS { unsigned long cr_end = mfctl(16); unsigned long tmp = cr_end - cr_start; @@ -404,8 +406,6 @@ resource_found: } 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 /* @@ -439,9 +439,9 @@ ccio_free_range(struct ioc *ioc, dma_addr_t iova, unsigned long pages_mapped) BUG_ON(pages_mapped > BITS_PER_LONG); DBG_RES("%s(): res_idx: %d pages_mapped %d\n", - __FUNCTION__, res_idx, pages_mapped); + __func__, res_idx, pages_mapped); -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->used_pages -= pages_mapped; #endif @@ -451,10 +451,10 @@ ccio_free_range(struct ioc *ioc, dma_addr_t iova, unsigned long pages_mapped) unsigned long mask = ~(~0UL >> pages_mapped); CCIO_FREE_MAPPINGS(ioc, res_idx, mask, 8); #else - CCIO_FREE_MAPPINGS(ioc, res_idx, 0xff, 8); + CCIO_FREE_MAPPINGS(ioc, res_idx, 0xffUL, 8); #endif } else if(pages_mapped <= 16) { - CCIO_FREE_MAPPINGS(ioc, res_idx, 0xffff, 16); + CCIO_FREE_MAPPINGS(ioc, res_idx, 0xffffUL, 16); } else if(pages_mapped <= 32) { CCIO_FREE_MAPPINGS(ioc, res_idx, ~(unsigned int)0, 32); #ifdef __LP64__ @@ -463,7 +463,7 @@ ccio_free_range(struct ioc *ioc, dma_addr_t iova, unsigned long pages_mapped) #endif } else { panic("%s:%s() Too many pages to unmap.\n", __FILE__, - __FUNCTION__); + __func__); } } @@ -485,7 +485,7 @@ typedef unsigned long space_t; ** 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. +** Apparently 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. @@ -553,7 +553,7 @@ static u32 hint_lookup[] = { * (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 +static void CCIO_INLINE ccio_io_pdir_entry(u64 *pdir_ptr, space_t sid, unsigned long vba, unsigned long hints) { @@ -595,7 +595,7 @@ ccio_io_pdir_entry(u64 *pdir_ptr, space_t sid, unsigned long vba, ** Grab virtual index [0:11] ** Deposit virt_idx bits into I/O PDIR word */ - asm volatile ("lci 0(%%sr1, %1), %0" : "=r" (ci) : "r" (vba)); + 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)); @@ -613,7 +613,7 @@ ccio_io_pdir_entry(u64 *pdir_ptr, space_t sid, unsigned long vba, ** the real mode coherence index generation of U2, the PDIR entry ** must be flushed to memory to retain coherence." */ - asm volatile("fdc 0(%0)" : : "r" (pdir_ptr)); + asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr)); asm volatile("sync"); } @@ -636,7 +636,7 @@ ccio_clear_io_tlb(struct ioc *ioc, dma_addr_t iovp, size_t byte_cnt) byte_cnt += chain_size; while(byte_cnt > chain_size) { - WRITE_U32(CMD_TLB_PURGE | iovp, &ioc->ioc_hpa->io_command); + WRITE_U32(CMD_TLB_PURGE | iovp, &ioc->ioc_regs->io_command); iovp += chain_size; byte_cnt -= chain_size; } @@ -651,7 +651,7 @@ ccio_clear_io_tlb(struct ioc *ioc, dma_addr_t iovp, size_t byte_cnt) * 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 + * FIXME: at some threshold 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 @@ -667,7 +667,7 @@ ccio_mark_invalid(struct ioc *ioc, dma_addr_t iova, size_t byte_cnt) size_t saved_byte_cnt; /* round up to nearest page size */ - saved_byte_cnt = byte_cnt = ROUNDUP(byte_cnt, IOVP_SIZE); + saved_byte_cnt = byte_cnt = ALIGN(byte_cnt, IOVP_SIZE); while(byte_cnt > 0) { /* invalidate one page at a time */ @@ -684,7 +684,7 @@ ccio_mark_invalid(struct ioc *ioc, dma_addr_t iova, size_t byte_cnt) ** Hopefully someone figures out how to patch (NOP) the ** FDC/SYNC out at boot time. */ - asm volatile("fdc 0(%0)" : : "r" (pdir_ptr[7])); + asm volatile("fdc %%r0(%0)" : : "r" (pdir_ptr[7])); iovp += IOVP_SIZE; byte_cnt -= IOVP_SIZE; @@ -750,21 +750,21 @@ ccio_map_single(struct device *dev, void *addr, size_t size, offset = ((unsigned long) addr) & ~IOVP_MASK; /* round up to nearest IOVP_SIZE */ - size = ROUNDUP(size + offset, IOVP_SIZE); + size = ALIGN(size + offset, IOVP_SIZE); spin_lock_irqsave(&ioc->res_lock, flags); -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->msingle_calls++; ioc->msingle_pages += size >> IOVP_SHIFT; #endif - idx = ccio_alloc_range(ioc, size); + idx = ccio_alloc_range(ioc, dev, 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); + __func__, 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)) @@ -809,15 +809,15 @@ ccio_unmap_single(struct device *dev, dma_addr_t iova, size_t size, ioc = GET_IOC(dev); DBG_RUN("%s() iovp 0x%lx/%x\n", - __FUNCTION__, (long)iova, size); + __func__, (long)iova, size); iova ^= offset; /* clear offset bits */ size += offset; - size = ROUNDUP(size, IOVP_SIZE); + size = ALIGN(size, IOVP_SIZE); spin_lock_irqsave(&ioc->res_lock, flags); -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->usingle_calls++; ioc->usingle_pages += size >> IOVP_SHIFT; #endif @@ -836,7 +836,7 @@ ccio_unmap_single(struct device *dev, dma_addr_t iova, size_t size, * This function implements the pci_alloc_consistent function. */ static void * -ccio_alloc_consistent(struct device *dev, size_t size, dma_addr_t *dma_handle, int flag) +ccio_alloc_consistent(struct device *dev, size_t size, dma_addr_t *dma_handle, gfp_t flag) { void *ret; #if 0 @@ -883,7 +883,7 @@ ccio_free_consistent(struct device *dev, size_t size, void *cpu_addr, */ #define PIDE_FLAG 0x80000000UL -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS #define IOMMU_MAP_STATS #endif #include "iommu-helpers.h" @@ -911,7 +911,7 @@ ccio_map_sg(struct device *dev, struct scatterlist *sglist, int nents, BUG_ON(!dev); ioc = GET_IOC(dev); - DBG_RUN_SG("%s() START %d entries\n", __FUNCTION__, nents); + DBG_RUN_SG("%s() START %d entries\n", __func__, nents); /* Fast path single entry scatterlists. */ if (nents == 1) { @@ -927,7 +927,7 @@ ccio_map_sg(struct device *dev, struct scatterlist *sglist, int nents, spin_lock_irqsave(&ioc->res_lock, flags); -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->msg_calls++; #endif @@ -939,7 +939,7 @@ ccio_map_sg(struct device *dev, struct scatterlist *sglist, int nents, ** 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); + coalesced = iommu_coalesce_chunks(ioc, dev, sglist, nents, ccio_alloc_range); /* ** Program the I/O Pdir @@ -955,7 +955,7 @@ ccio_map_sg(struct device *dev, struct scatterlist *sglist, int nents, BUG_ON(coalesced != filled); - DBG_RUN_SG("%s() DONE %d mappings\n", __FUNCTION__, filled); + DBG_RUN_SG("%s() DONE %d mappings\n", __func__, filled); for (i = 0; i < filled; i++) current_len += sg_dma_len(sglist + i); @@ -984,15 +984,15 @@ ccio_unmap_sg(struct device *dev, struct scatterlist *sglist, int nents, ioc = GET_IOC(dev); DBG_RUN_SG("%s() START %d entries, %08lx,%x\n", - __FUNCTION__, nents, sg_virt_addr(sglist), sglist->length); + __func__, nents, sg_virt_addr(sglist), sglist->length); -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->usg_calls++; #endif while(sg_dma_len(sglist) && nents--) { -#ifdef CCIO_MAP_STATS +#ifdef CCIO_COLLECT_STATS ioc->usg_pages += sg_dma_len(sglist) >> PAGE_SHIFT; #endif ccio_unmap_single(dev, sg_dma_address(sglist), @@ -1000,7 +1000,7 @@ ccio_unmap_sg(struct device *dev, struct scatterlist *sglist, int nents, ++sglist; } - DBG_RUN_SG("%s() DONE (nents %d)\n", __FUNCTION__, nents); + DBG_RUN_SG("%s() DONE (nents %d)\n", __func__, nents); } static struct hppa_dma_ops ccio_ops = { @@ -1019,63 +1019,36 @@ static struct hppa_dma_ops ccio_ops = { }; #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) +static int ccio_proc_info(struct seq_file *m, void *p) { - int max = count; - char tmp[80]; /* width of an ANSI-standard terminal */ + int len = 0; struct ioc *ioc = ioc_list; while (ioc != NULL) { unsigned int total_pages = ioc->res_size << 3; +#ifdef CCIO_COLLECT_STATS unsigned long avg = 0, min, max; - int j, len; + int j; +#endif - len = sprintf(tmp, "%s\n", ioc->name); - if (proc_append(tmp, len, &buf, &offset, &count)) - break; + len += seq_printf(m, "%s\n", ioc->name); - len = sprintf(tmp, "Cujo 2.0 bug : %s\n", - (ioc->cujo20_bug ? "yes" : "no")); - if (proc_append(tmp, len, &buf, &offset, &count)) - break; + len += seq_printf(m, "Cujo 2.0 bug : %s\n", + (ioc->cujo20_bug ? "yes" : "no")); - 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; + len += seq_printf(m, "IO PDIR size : %d bytes (%d entries)\n", + total_pages * 8, total_pages); + +#ifdef CCIO_COLLECT_STATS + len += seq_printf(m, "IO PDIR entries : %ld free %ld used (%d%%)\n", + total_pages - ioc->used_pages, ioc->used_pages, + (int)(ioc->used_pages * 100 / total_pages)); #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 + + len += seq_printf(m, "Resource bitmap : %d bytes (%d pages)\n", + ioc->res_size, total_pages); + +#ifdef CCIO_COLLECT_STATS min = max = ioc->avg_search[0]; for(j = 0; j < CCIO_SEARCH_SAMPLE; ++j) { avg += ioc->avg_search[j]; @@ -1085,71 +1058,83 @@ static int ccio_proc_info(char *buf, char **start, off_t offset, int count, 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; - + len += seq_printf(m, " Bitmap search : %ld/%ld/%ld (min/avg/max CPU Cycles)\n", + min, avg, max); + + len += seq_printf(m, "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)); /* 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 += seq_printf(m, "pci_unmap_single: %8ld calls %8ld pages (avg %d/1000)\n", + min, max, (int)((max * 1000)/min)); - 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 */ + len += seq_printf(m, "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)); + + len += seq_printf(m, "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)); +#endif /* CCIO_COLLECT_STATS */ + ioc = ioc->next; } - if (count == 0) { - *eof = 1; - } - return (max - count); + return 0; } -static int ccio_resource_map(char *buf, char **start, off_t offset, int len, - int *eof, void *data) +static int ccio_proc_info_open(struct inode *inode, struct file *file) +{ + return single_open(file, &ccio_proc_info, NULL); +} + +static const struct file_operations ccio_proc_info_fops = { + .owner = THIS_MODULE, + .open = ccio_proc_info_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +static int ccio_proc_bitmap_info(struct seq_file *m, void *p) { + int len = 0; 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); + len += seq_puts(m, "\n "); + len += seq_printf(m, "%08x", *res_ptr); res_ptr++; } - strcat(buf, "\n\n"); + len += seq_puts(m, "\n\n"); ioc = ioc->next; break; /* XXX - remove me */ } - return strlen(buf); + return 0; +} + +static int ccio_proc_bitmap_open(struct inode *inode, struct file *file) +{ + return single_open(file, &ccio_proc_bitmap_info, NULL); } -#endif + +static const struct file_operations ccio_proc_bitmap_fops = { + .owner = THIS_MODULE, + .open = ccio_proc_bitmap_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; +#endif /* CONFIG_PROC_FS */ /** * ccio_find_ioc - Find the ioc in the ioc_list @@ -1230,7 +1215,7 @@ 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__); + panic("%s() : Can't determine I/O TLB size.\n", __func__); } return (1 << dev->spa_shift); } @@ -1242,7 +1227,7 @@ ccio_get_iotlb_size(struct parisc_device *dev) #endif /* 0 */ /* We *can't* support JAVA (T600). Venture there at your own risk. */ -static struct parisc_device_id ccio_tbl[] = { +static const 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, } @@ -1251,16 +1236,16 @@ static struct parisc_device_id ccio_tbl[] = { static int ccio_probe(struct parisc_device *dev); static struct parisc_driver ccio_driver = { - .name = "U2:Uturn", + .name = "ccio", .id_table = ccio_tbl, .probe = ccio_probe, }; /** - * ccio_ioc_init - Initalize the I/O Controller + * ccio_ioc_init - Initialize the I/O Controller * @ioc: The I/O Controller. * - * Initalize the I/O Controller which includes setting up the + * Initialize 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. */ @@ -1282,7 +1267,7 @@ ccio_ioc_init(struct ioc *ioc) ** Hot-Plug/Removal of PCI cards. (aka PCI OLARD). */ - iova_space_size = (u32) (num_physpages / count_parisc_driver(&ccio_driver)); + iova_space_size = (u32) (totalram_pages / count_parisc_driver(&ccio_driver)); /* limit IOVA space size to 1MB-1GB */ @@ -1314,38 +1299,35 @@ ccio_ioc_init(struct ioc *ioc) ioc->pdir_size = (iova_space_size / IOVP_SIZE) * sizeof(u64); - BUG_ON(ioc->pdir_size >= 4 * 1024 * 1024); /* max pdir size < 4MB */ + 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%lx mem %luMB IOV %dMB (%d bits)\n", - __FUNCTION__, - ioc->ioc_hpa, - (unsigned long) num_physpages >> (20 - PAGE_SHIFT), + DBG_INIT("%s() hpa 0x%p mem %luMB IOV %dMB (%d bits)\n", + __func__, ioc->ioc_regs, + (unsigned long) totalram_pages >> (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:%s() could not allocate I/O Page Table\n", __FILE__, - __FUNCTION__); + panic("%s() could not allocate I/O Page Table\n", __func__); } 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", 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); + DBG_INIT("%s() res_size 0x%x\n", __func__, ioc->res_size); ioc->res_map = (u8 *)__get_free_pages(GFP_KERNEL, get_order(ioc->res_size)); if(NULL == ioc->res_map) { - panic("%s:%s() could not allocate resource map\n", __FILE__, - __FUNCTION__); + panic("%s() could not allocate resource map\n", __func__); } memset(ioc->res_map, 0, ioc->res_size); @@ -1366,44 +1348,58 @@ ccio_ioc_init(struct ioc *ioc) ** Initialize IOA hardware */ WRITE_U32(CCIO_CHAINID_MASK << ioc->chainid_shift, - &ioc->ioc_hpa->io_chain_id_mask); + &ioc->ioc_regs->io_chain_id_mask); WRITE_U32(virt_to_phys(ioc->pdir_base), - &ioc->ioc_hpa->io_pdir_base); + &ioc->ioc_regs->io_pdir_base); /* ** Go to "Virtual Mode" */ - WRITE_U32(IOA_NORMAL_MODE, &ioc->ioc_hpa->io_control); + 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_hpa->io_tlb_entry_m); - WRITE_U32(0, &ioc->ioc_hpa->io_tlb_entry_l); + 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_hpa->io_command); + &ioc->ioc_regs->io_command); } } -static void -ccio_init_resource(struct resource *res, char *name, unsigned long ioaddr) +static void __init +ccio_init_resource(struct resource *res, char *name, void __iomem *ioaddr) { int result; res->parent = NULL; res->flags = IORESOURCE_MEM; - res->start = (unsigned long)(signed) __raw_readl(ioaddr) << 16; - res->end = (unsigned long)(signed) (__raw_readl(ioaddr + 4) << 16) - 1; + /* + * 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; - result = request_resource(&iomem_resource, res); + + /* 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", - __FILE__, res->start, res->end); + printk(KERN_ERR "%s() failed to claim CCIO bus address space (%08lx,%08lx)\n", + __func__, (unsigned long)res->start, (unsigned long)res->end); } } @@ -1412,11 +1408,10 @@ static void __init ccio_init_resources(struct ioc *ioc) struct resource *res = ioc->mmio_region; char *name = kmalloc(14, GFP_KERNEL); - sprintf(name, "GSC Bus [%d/]", ioc->hw_path); + snprintf(name, 14, "GSC Bus [%d/]", ioc->hw_path); - ccio_init_resource(res, name, (unsigned long)&ioc->ioc_hpa->io_io_low); - ccio_init_resource(res + 1, name, - (unsigned long)&ioc->ioc_hpa->io_io_low_hv); + 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, @@ -1427,7 +1422,12 @@ static int new_ioc_area(struct resource *res, unsigned long size, res->start = (max - size + 1) &~ (align - 1); res->end = res->start + size; - if (!request_resource(&iomem_resource, res)) + + /* 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); @@ -1486,15 +1486,15 @@ int ccio_allocate_resource(const struct parisc_device *dev, if (!expand_ioc_area(parent, size, min, max, align)) { __raw_writel(((parent->start)>>16) | 0xffff0000, - (unsigned long)&(ioc->ioc_hpa->io_io_low)); + &ioc->ioc_regs->io_io_low); __raw_writel(((parent->end)>>16) | 0xffff0000, - (unsigned long)&(ioc->ioc_hpa->io_io_high)); + &ioc->ioc_regs->io_io_high); } else if (!expand_ioc_area(parent + 1, size, min, max, align)) { parent++; __raw_writel(((parent->start)>>16) | 0xffff0000, - (unsigned long)&(ioc->ioc_hpa->io_io_low_hv)); + &ioc->ioc_regs->io_io_low_hv); __raw_writel(((parent->end)>>16) | 0xffff0000, - (unsigned long)&(ioc->ioc_hpa->io_io_high_hv)); + &ioc->ioc_regs->io_io_high_hv); } else { return -EBUSY; } @@ -1521,7 +1521,12 @@ int ccio_request_resource(const struct parisc_device *dev, return -EBUSY; } - return request_resource(parent, res); + /* "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); } /** @@ -1532,21 +1537,21 @@ int ccio_request_resource(const struct parisc_device *dev, * If so, initialize the chip and tell other partners in crime they * have work to do. */ -static int ccio_probe(struct parisc_device *dev) +static int __init ccio_probe(struct parisc_device *dev) { int i; struct ioc *ioc, **ioc_p = &ioc_list; - - ioc = kmalloc(sizeof(struct ioc), GFP_KERNEL); + + ioc = kzalloc(sizeof(struct ioc), GFP_KERNEL); if (ioc == NULL) { printk(KERN_ERR MODULE_NAME ": memory allocation failure\n"); return 1; } - memset(ioc, 0, sizeof(struct ioc)); ioc->name = dev->id.hversion == U2_IOA_RUNWAY ? "U2" : "UTurn"; - printk(KERN_INFO "Found %s at 0x%lx\n", ioc->name, dev->hpa); + printk(KERN_INFO "Found %s at 0x%lx\n", ioc->name, + (unsigned long)dev->hpa.start); for (i = 0; i < ioc_count; i++) { ioc_p = &(*ioc_p)->next; @@ -1554,35 +1559,32 @@ static int ccio_probe(struct parisc_device *dev) *ioc_p = ioc; ioc->hw_path = dev->hw_path; - ioc->ioc_hpa = (struct ioa_registers *)dev->hpa; + ioc->ioc_regs = ioremap_nocache(dev->hpa.start, 4096); ccio_ioc_init(ioc); ccio_init_resources(ioc); hppa_dma_ops = &ccio_ops; - dev->dev.platform_data = kmalloc(sizeof(struct pci_hba_data), GFP_KERNEL); + 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; - +#ifdef CONFIG_PROC_FS 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); + proc_create(MODULE_NAME, 0, proc_runway_root, + &ccio_proc_info_fops); + proc_create(MODULE_NAME"-bitmap", 0, proc_runway_root, + &ccio_proc_bitmap_fops); } - +#endif 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. + * ccio_init - ccio initialization procedure. * * Register this driver. */ |