Skip to content

Commit

Permalink
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Browse files Browse the repository at this point in the history
Pull sparc fixes from David Miller
 "Unfortunately, I brown paper bagged the generic iommu pool allocator
  by applying the wrong revision of the patch series.

  This reverts the bad one, and puts the right one in"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  iommu-common: Fix PARISC compile-time warnings
  sparc: Make LDC use common iommu poll management functions
  sparc: Make sparc64 use scalable lib/iommu-common.c functions
  Break up monolithic iommu table/lock into finer graularity pools and lock
  sparc: Revert generic IOMMU allocator.
  • Loading branch information
torvalds committed Apr 18, 2015
2 parents dba94f2 + ccb3018 commit 64fb1d0
Show file tree
Hide file tree
Showing 6 changed files with 202 additions and 230 deletions.
2 changes: 1 addition & 1 deletion arch/sparc/include/asm/iommu_64.h
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ struct iommu_arena {
};

struct iommu {
struct iommu_table tbl;
struct iommu_map_table tbl;
spinlock_t lock;
u32 dma_addr_mask;
iopte_t *page_table;
Expand Down
78 changes: 27 additions & 51 deletions arch/sparc/kernel/iommu.c
Original file line number Diff line number Diff line change
Expand Up @@ -13,15 +13,12 @@
#include <linux/errno.h>
#include <linux/iommu-helper.h>
#include <linux/bitmap.h>
#include <linux/hash.h>
#include <linux/iommu-common.h>

#ifdef CONFIG_PCI
#include <linux/pci.h>
#endif

static DEFINE_PER_CPU(unsigned int, iommu_pool_hash);

#include <asm/iommu.h>

#include "iommu_common.h"
Expand Down Expand Up @@ -49,9 +46,9 @@ static DEFINE_PER_CPU(unsigned int, iommu_pool_hash);
"i" (ASI_PHYS_BYPASS_EC_E))

/* Must be invoked under the IOMMU lock. */
static void iommu_flushall(struct iommu_table *iommu_table)
static void iommu_flushall(struct iommu_map_table *iommu_map_table)
{
struct iommu *iommu = container_of(iommu_table, struct iommu, tbl);
struct iommu *iommu = container_of(iommu_map_table, struct iommu, tbl);
if (iommu->iommu_flushinv) {
iommu_write(iommu->iommu_flushinv, ~(u64)0);
} else {
Expand Down Expand Up @@ -92,23 +89,6 @@ static inline void iopte_make_dummy(struct iommu *iommu, iopte_t *iopte)
iopte_val(*iopte) = val;
}

static struct iommu_tbl_ops iommu_sparc_ops = {
.reset = iommu_flushall
};

static void setup_iommu_pool_hash(void)
{
unsigned int i;
static bool do_once;

if (do_once)
return;
do_once = true;
for_each_possible_cpu(i)
per_cpu(iommu_pool_hash, i) = hash_32(i, IOMMU_POOL_HASHBITS);
}


int iommu_table_init(struct iommu *iommu, int tsbsize,
u32 dma_offset, u32 dma_addr_mask,
int numa_node)
Expand All @@ -121,7 +101,7 @@ int iommu_table_init(struct iommu *iommu, int tsbsize,
/* Setup initial software IOMMU state. */
spin_lock_init(&iommu->lock);
iommu->ctx_lowest_free = 1;
iommu->tbl.page_table_map_base = dma_offset;
iommu->tbl.table_map_base = dma_offset;
iommu->dma_addr_mask = dma_addr_mask;

/* Allocate and initialize the free area map. */
Expand All @@ -131,12 +111,10 @@ int iommu_table_init(struct iommu *iommu, int tsbsize,
if (!iommu->tbl.map)
return -ENOMEM;
memset(iommu->tbl.map, 0, sz);
if (tlb_type != hypervisor)
iommu_sparc_ops.reset = NULL; /* not needed on on sun4v */

setup_iommu_pool_hash();
iommu_tbl_pool_init(&iommu->tbl, num_tsb_entries, IO_PAGE_SHIFT,
&iommu_sparc_ops, false, 1);
(tlb_type != hypervisor ? iommu_flushall : NULL),
false, 1, false);

/* Allocate and initialize the dummy page which we
* set inactive IO PTEs to point to.
Expand Down Expand Up @@ -182,7 +160,7 @@ static inline iopte_t *alloc_npages(struct device *dev,
unsigned long entry;

entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages, NULL,
__this_cpu_read(iommu_pool_hash));
(unsigned long)(-1), 0);
if (unlikely(entry == DMA_ERROR_CODE))
return NULL;

Expand Down Expand Up @@ -249,7 +227,7 @@ static void *dma_4u_alloc_coherent(struct device *dev, size_t size,
return NULL;
}

*dma_addrp = (iommu->tbl.page_table_map_base +
*dma_addrp = (iommu->tbl.table_map_base +
((iopte - iommu->page_table) << IO_PAGE_SHIFT));
ret = (void *) first_page;
npages = size >> IO_PAGE_SHIFT;
Expand All @@ -275,7 +253,7 @@ static void dma_4u_free_coherent(struct device *dev, size_t size,
npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT;
iommu = dev->archdata.iommu;

iommu_tbl_range_free(&iommu->tbl, dvma, npages, false, NULL);
iommu_tbl_range_free(&iommu->tbl, dvma, npages, DMA_ERROR_CODE);

order = get_order(size);
if (order < 10)
Expand Down Expand Up @@ -315,7 +293,7 @@ static dma_addr_t dma_4u_map_page(struct device *dev, struct page *page,
if (unlikely(!base))
goto bad;

bus_addr = (iommu->tbl.page_table_map_base +
bus_addr = (iommu->tbl.table_map_base +
((base - iommu->page_table) << IO_PAGE_SHIFT));
ret = bus_addr | (oaddr & ~IO_PAGE_MASK);
base_paddr = __pa(oaddr & IO_PAGE_MASK);
Expand Down Expand Up @@ -426,7 +404,7 @@ static void dma_4u_unmap_page(struct device *dev, dma_addr_t bus_addr,
npages = IO_PAGE_ALIGN(bus_addr + sz) - (bus_addr & IO_PAGE_MASK);
npages >>= IO_PAGE_SHIFT;
base = iommu->page_table +
((bus_addr - iommu->tbl.page_table_map_base) >> IO_PAGE_SHIFT);
((bus_addr - iommu->tbl.table_map_base) >> IO_PAGE_SHIFT);
bus_addr &= IO_PAGE_MASK;

spin_lock_irqsave(&iommu->lock, flags);
Expand All @@ -448,8 +426,7 @@ static void dma_4u_unmap_page(struct device *dev, dma_addr_t bus_addr,
iommu_free_ctx(iommu, ctx);
spin_unlock_irqrestore(&iommu->lock, flags);

iommu_tbl_range_free(&iommu->tbl, bus_addr, npages,
false, NULL);
iommu_tbl_range_free(&iommu->tbl, bus_addr, npages, DMA_ERROR_CODE);
}

static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
Expand Down Expand Up @@ -497,7 +474,7 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
max_seg_size = dma_get_max_seg_size(dev);
seg_boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
IO_PAGE_SIZE) >> IO_PAGE_SHIFT;
base_shift = iommu->tbl.page_table_map_base >> IO_PAGE_SHIFT;
base_shift = iommu->tbl.table_map_base >> IO_PAGE_SHIFT;
for_each_sg(sglist, s, nelems, i) {
unsigned long paddr, npages, entry, out_entry = 0, slen;
iopte_t *base;
Expand All @@ -511,8 +488,8 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
/* Allocate iommu entries for that segment */
paddr = (unsigned long) SG_ENT_PHYS_ADDRESS(s);
npages = iommu_num_pages(paddr, slen, IO_PAGE_SIZE);
entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages, &handle,
__this_cpu_read(iommu_pool_hash));
entry = iommu_tbl_range_alloc(dev, &iommu->tbl, npages,
&handle, (unsigned long)(-1), 0);

/* Handle failure */
if (unlikely(entry == DMA_ERROR_CODE)) {
Expand All @@ -525,7 +502,7 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
base = iommu->page_table + entry;

/* Convert entry to a dma_addr_t */
dma_addr = iommu->tbl.page_table_map_base +
dma_addr = iommu->tbl.table_map_base +
(entry << IO_PAGE_SHIFT);
dma_addr |= (s->offset & ~IO_PAGE_MASK);

Expand Down Expand Up @@ -586,15 +563,15 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
npages = iommu_num_pages(s->dma_address, s->dma_length,
IO_PAGE_SIZE);

entry = (vaddr - iommu->tbl.page_table_map_base)
entry = (vaddr - iommu->tbl.table_map_base)
>> IO_PAGE_SHIFT;
base = iommu->page_table + entry;

for (j = 0; j < npages; j++)
iopte_make_dummy(iommu, base + j);

iommu_tbl_range_free(&iommu->tbl, vaddr, npages,
false, NULL);
DMA_ERROR_CODE);

s->dma_address = DMA_ERROR_CODE;
s->dma_length = 0;
Expand All @@ -610,19 +587,18 @@ static int dma_4u_map_sg(struct device *dev, struct scatterlist *sglist,
/* If contexts are being used, they are the same in all of the mappings
* we make for a particular SG.
*/
static unsigned long fetch_sg_ctx(struct iommu *iommu,
struct scatterlist *sg)
static unsigned long fetch_sg_ctx(struct iommu *iommu, struct scatterlist *sg)
{
unsigned long ctx = 0;

if (iommu->iommu_ctxflush) {
iopte_t *base;
u32 bus_addr;
struct iommu_table *tbl = &iommu->tbl;
struct iommu_map_table *tbl = &iommu->tbl;

bus_addr = sg->dma_address & IO_PAGE_MASK;
base = iommu->page_table +
((bus_addr - tbl->page_table_map_base) >> IO_PAGE_SHIFT);
((bus_addr - tbl->table_map_base) >> IO_PAGE_SHIFT);

ctx = (iopte_val(*base) & IOPTE_CONTEXT) >> 47UL;
}
Expand Down Expand Up @@ -659,7 +635,7 @@ static void dma_4u_unmap_sg(struct device *dev, struct scatterlist *sglist,
break;
npages = iommu_num_pages(dma_handle, len, IO_PAGE_SIZE);

entry = ((dma_handle - iommu->tbl.page_table_map_base)
entry = ((dma_handle - iommu->tbl.table_map_base)
>> IO_PAGE_SHIFT);
base = iommu->page_table + entry;

Expand All @@ -671,8 +647,8 @@ static void dma_4u_unmap_sg(struct device *dev, struct scatterlist *sglist,
for (i = 0; i < npages; i++)
iopte_make_dummy(iommu, base + i);

iommu_tbl_range_free(&iommu->tbl, dma_handle, npages, false,
NULL);
iommu_tbl_range_free(&iommu->tbl, dma_handle, npages,
DMA_ERROR_CODE);
sg = sg_next(sg);
}

Expand Down Expand Up @@ -706,10 +682,10 @@ static void dma_4u_sync_single_for_cpu(struct device *dev,
if (iommu->iommu_ctxflush &&
strbuf->strbuf_ctxflush) {
iopte_t *iopte;
struct iommu_table *tbl = &iommu->tbl;
struct iommu_map_table *tbl = &iommu->tbl;

iopte = iommu->page_table +
((bus_addr - tbl->page_table_map_base)>>IO_PAGE_SHIFT);
((bus_addr - tbl->table_map_base)>>IO_PAGE_SHIFT);
ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL;
}

Expand Down Expand Up @@ -742,10 +718,10 @@ static void dma_4u_sync_sg_for_cpu(struct device *dev,
if (iommu->iommu_ctxflush &&
strbuf->strbuf_ctxflush) {
iopte_t *iopte;
struct iommu_table *tbl = &iommu->tbl;
struct iommu_map_table *tbl = &iommu->tbl;

iopte = iommu->page_table + ((sglist[0].dma_address -
tbl->page_table_map_base) >> IO_PAGE_SHIFT);
tbl->table_map_base) >> IO_PAGE_SHIFT);
ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL;
}

Expand Down
60 changes: 14 additions & 46 deletions arch/sparc/kernel/ldc.c
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,6 @@
#include <linux/list.h>
#include <linux/init.h>
#include <linux/bitmap.h>
#include <linux/hash.h>
#include <linux/iommu-common.h>

#include <asm/hypervisor.h>
Expand All @@ -32,7 +31,6 @@
#define COOKIE_PGSZ_CODE 0xf000000000000000ULL
#define COOKIE_PGSZ_CODE_SHIFT 60ULL

static DEFINE_PER_CPU(unsigned int, ldc_pool_hash);

static char version[] =
DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
Expand Down Expand Up @@ -108,7 +106,7 @@ struct ldc_iommu {
/* Protects ldc_unmap. */
spinlock_t lock;
struct ldc_mtable_entry *page_table;
struct iommu_table iommu_table;
struct iommu_map_table iommu_map_table;
};

struct ldc_channel {
Expand Down Expand Up @@ -1015,18 +1013,9 @@ static unsigned long ldc_cookie_to_index(u64 cookie, void *arg)
return (cookie >> (13ULL + (szcode * 3ULL)));
}

struct ldc_demap_arg {
struct ldc_iommu *ldc_iommu;
u64 cookie;
unsigned long id;
};

static void ldc_demap(void *arg, unsigned long entry, unsigned long npages)
static void ldc_demap(struct ldc_iommu *iommu, unsigned long id, u64 cookie,
unsigned long entry, unsigned long npages)
{
struct ldc_demap_arg *ldc_demap_arg = arg;
struct ldc_iommu *iommu = ldc_demap_arg->ldc_iommu;
unsigned long id = ldc_demap_arg->id;
u64 cookie = ldc_demap_arg->cookie;
struct ldc_mtable_entry *base;
unsigned long i, shift;

Expand All @@ -1043,36 +1032,17 @@ static void ldc_demap(void *arg, unsigned long entry, unsigned long npages)
/* XXX Make this configurable... XXX */
#define LDC_IOTABLE_SIZE (8 * 1024)

struct iommu_tbl_ops ldc_iommu_ops = {
.cookie_to_index = ldc_cookie_to_index,
.demap = ldc_demap,
};

static void setup_ldc_pool_hash(void)
{
unsigned int i;
static bool do_once;

if (do_once)
return;
do_once = true;
for_each_possible_cpu(i)
per_cpu(ldc_pool_hash, i) = hash_32(i, IOMMU_POOL_HASHBITS);
}


static int ldc_iommu_init(const char *name, struct ldc_channel *lp)
{
unsigned long sz, num_tsb_entries, tsbsize, order;
struct ldc_iommu *ldc_iommu = &lp->iommu;
struct iommu_table *iommu = &ldc_iommu->iommu_table;
struct iommu_map_table *iommu = &ldc_iommu->iommu_map_table;
struct ldc_mtable_entry *table;
unsigned long hv_err;
int err;

num_tsb_entries = LDC_IOTABLE_SIZE;
tsbsize = num_tsb_entries * sizeof(struct ldc_mtable_entry);
setup_ldc_pool_hash();
spin_lock_init(&ldc_iommu->lock);

sz = num_tsb_entries / 8;
Expand All @@ -1083,7 +1053,9 @@ static int ldc_iommu_init(const char *name, struct ldc_channel *lp)
return -ENOMEM;
}
iommu_tbl_pool_init(iommu, num_tsb_entries, PAGE_SHIFT,
&ldc_iommu_ops, false, 1);
NULL, false /* no large pool */,
1 /* npools */,
true /* skip span boundary check */);

order = get_order(tsbsize);

Expand Down Expand Up @@ -1122,7 +1094,7 @@ static int ldc_iommu_init(const char *name, struct ldc_channel *lp)
static void ldc_iommu_release(struct ldc_channel *lp)
{
struct ldc_iommu *ldc_iommu = &lp->iommu;
struct iommu_table *iommu = &ldc_iommu->iommu_table;
struct iommu_map_table *iommu = &ldc_iommu->iommu_map_table;
unsigned long num_tsb_entries, tsbsize, order;

(void) sun4v_ldc_set_map_table(lp->id, 0, 0);
Expand Down Expand Up @@ -1979,8 +1951,8 @@ static struct ldc_mtable_entry *alloc_npages(struct ldc_iommu *iommu,
{
long entry;

entry = iommu_tbl_range_alloc(NULL, &iommu->iommu_table, npages,
NULL, __this_cpu_read(ldc_pool_hash));
entry = iommu_tbl_range_alloc(NULL, &iommu->iommu_map_table,
npages, NULL, (unsigned long)-1, 0);
if (unlikely(entry < 0))
return NULL;

Expand Down Expand Up @@ -2191,17 +2163,13 @@ EXPORT_SYMBOL(ldc_map_single);
static void free_npages(unsigned long id, struct ldc_iommu *iommu,
u64 cookie, u64 size)
{
unsigned long npages;
struct ldc_demap_arg demap_arg;

demap_arg.ldc_iommu = iommu;
demap_arg.cookie = cookie;
demap_arg.id = id;
unsigned long npages, entry;

npages = PAGE_ALIGN(((cookie & ~PAGE_MASK) + size)) >> PAGE_SHIFT;
iommu_tbl_range_free(&iommu->iommu_table, cookie, npages, true,
&demap_arg);

entry = ldc_cookie_to_index(cookie, iommu);
ldc_demap(iommu, id, cookie, entry, npages);
iommu_tbl_range_free(&iommu->iommu_map_table, cookie, npages, entry);
}

void ldc_unmap(struct ldc_channel *lp, struct ldc_trans_cookie *cookies,
Expand Down
Loading

0 comments on commit 64fb1d0

Please sign in to comment.