bp_map.c revision 36945f796054e8cb46d88ec0a84213123cf2b036
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#pragma ident "%Z%%M% %I% %E% SMI"
#include <sys/types.h>
#include <sys/sysmacros.h>
#include <sys/systm.h>
#include <sys/mman.h>
#include <sys/buf.h>
#include <sys/vmem.h>
#include <sys/cmn_err.h>
#include <sys/debug.h>
#include <sys/machparam.h>
#include <vm/page.h>
#include <vm/seg_kmem.h>
#include <vm/seg_kpm.h>
#ifdef __sparc
#include <sys/cpu_module.h>
#define BP_FLUSH(addr, size) flush_instr_mem((void *)addr, size);
#else
#define BP_FLUSH(addr, size)
#endif
int bp_force_copy = 0;
typedef enum {
BP_COPYIN = 0,
BP_COPYOUT = 1
} bp_copydir_t;
static int bp_copy_common(bp_copydir_t dir, struct buf *bp, void *driverbuf,
offset_t offset, size_t size);
static vmem_t *bp_map_arena;
static size_t bp_align;
static uint_t bp_devload_flags = PROT_READ | PROT_WRITE | HAT_NOSYNC;
int bp_max_cache = 1 << 17; /* 128K default; tunable */
int bp_mapin_kpm_enable = 1; /* enable default; tunable */
static void *
bp_vmem_alloc(vmem_t *vmp, size_t size, int vmflag)
{
return (vmem_xalloc(vmp, size, bp_align, 0, 0, NULL, NULL, vmflag));
}
void
bp_init(size_t align, uint_t devload_flags)
{
bp_align = MAX(align, PAGESIZE);
bp_devload_flags |= devload_flags;
if (bp_align <= bp_max_cache)
bp_map_arena = vmem_create("bp_map", NULL, 0, bp_align,
bp_vmem_alloc, vmem_free, heap_arena,
MIN(8 * bp_align, bp_max_cache), VM_SLEEP);
}
/*
* common routine so can be called with/without VM_SLEEP
*/
void *
bp_mapin_common(struct buf *bp, int flag)
{
struct as *as;
pfn_t pfnum;
page_t *pp;
page_t **pplist;
caddr_t kaddr;
caddr_t addr;
uintptr_t off;
size_t size;
pgcnt_t npages;
int color;
/* return if already mapped in, no pageio/physio, or physio to kas */
if ((bp->b_flags & B_REMAPPED) ||
!(bp->b_flags & (B_PAGEIO | B_PHYS)) ||
(((bp->b_flags & (B_PAGEIO | B_PHYS)) == B_PHYS) &&
((bp->b_proc == NULL) || (bp->b_proc->p_as == &kas))))
return (bp->b_un.b_addr);
ASSERT((bp->b_flags & (B_PAGEIO | B_PHYS)) != (B_PAGEIO | B_PHYS));
addr = (caddr_t)bp->b_un.b_addr;
off = (uintptr_t)addr & PAGEOFFSET;
size = P2ROUNDUP(bp->b_bcount + off, PAGESIZE);
npages = btop(size);
/* Fastpath single page IO to locked memory by using kpm. */
if ((bp->b_flags & (B_SHADOW | B_PAGEIO)) && (npages == 1) &&
kpm_enable && bp_mapin_kpm_enable) {
if (bp->b_flags & B_SHADOW)
pp = *bp->b_shadow;
else
pp = bp->b_pages;
kaddr = hat_kpm_mapin(pp, NULL);
bp->b_un.b_addr = kaddr + off;
bp->b_flags |= B_REMAPPED;
return (bp->b_un.b_addr);
}
/*
* Allocate kernel virtual space for remapping.
*/
color = bp_color(bp);
ASSERT(color < bp_align);
if (bp_map_arena != NULL) {
kaddr = (caddr_t)vmem_alloc(bp_map_arena,
P2ROUNDUP(color + size, bp_align), flag);
if (kaddr == NULL)
return (NULL);
kaddr += color;
} else {
kaddr = vmem_xalloc(heap_arena, size, bp_align, color,
0, NULL, NULL, flag);
if (kaddr == NULL)
return (NULL);
}
ASSERT(P2PHASE((uintptr_t)kaddr, bp_align) == color);
/*
* Map bp into the virtual space we just allocated.
*/
if (bp->b_flags & B_PAGEIO) {
pp = bp->b_pages;
pplist = NULL;
} else if (bp->b_flags & B_SHADOW) {
pp = NULL;
pplist = bp->b_shadow;
} else {
pp = NULL;
pplist = NULL;
if (bp->b_proc == NULL || (as = bp->b_proc->p_as) == NULL)
as = &kas;
}
bp->b_flags |= B_REMAPPED;
bp->b_un.b_addr = kaddr + off;
while (npages-- != 0) {
if (pp) {
pfnum = pp->p_pagenum;
pp = pp->p_next;
} else if (pplist == NULL) {
pfnum = hat_getpfnum(as->a_hat,
(caddr_t)((uintptr_t)addr & MMU_PAGEMASK));
if (pfnum == PFN_INVALID)
panic("bp_mapin_common: hat_getpfnum for"
" addr %p failed\n", (void *)addr);
addr += PAGESIZE;
} else {
pfnum = (*pplist)->p_pagenum;
pplist++;
}
hat_devload(kas.a_hat, kaddr, PAGESIZE, pfnum,
bp_devload_flags, HAT_LOAD_LOCK);
kaddr += PAGESIZE;
}
return (bp->b_un.b_addr);
}
/*
* Convert bp for pageio/physio to a kernel addressable location.
*/
void
bp_mapin(struct buf *bp)
{
(void) bp_mapin_common(bp, VM_SLEEP);
}
/*
* Release all the resources associated with a previous bp_mapin() call.
*/
void
bp_mapout(struct buf *bp)
{
caddr_t addr;
uintptr_t off;
uintptr_t base;
uintptr_t color;
size_t size;
pgcnt_t npages;
page_t *pp;
if ((bp->b_flags & B_REMAPPED) == 0)
return;
addr = bp->b_un.b_addr;
off = (uintptr_t)addr & PAGEOFFSET;
size = P2ROUNDUP(bp->b_bcount + off, PAGESIZE);
npages = btop(size);
bp->b_un.b_addr = (caddr_t)off; /* debugging aid */
if ((bp->b_flags & (B_SHADOW | B_PAGEIO)) && (npages == 1) &&
kpm_enable && bp_mapin_kpm_enable) {
if (bp->b_flags & B_SHADOW)
pp = *bp->b_shadow;
else
pp = bp->b_pages;
addr = (caddr_t)((uintptr_t)addr & MMU_PAGEMASK);
hat_kpm_mapout(pp, NULL, addr);
bp->b_flags &= ~B_REMAPPED;
return;
}
base = (uintptr_t)addr & MMU_PAGEMASK;
BP_FLUSH(base, size);
hat_unload(kas.a_hat, (void *)base, size,
HAT_UNLOAD_NOSYNC | HAT_UNLOAD_UNLOCK);
if (bp_map_arena != NULL) {
color = P2PHASE(base, bp_align);
vmem_free(bp_map_arena, (void *)(base - color),
P2ROUNDUP(color + size, bp_align));
} else
vmem_free(heap_arena, (void *)base, size);
bp->b_flags &= ~B_REMAPPED;
}
/*
* copy data from a KVA into a buf_t which may not be mapped in. offset
* is relative to the buf_t only.
*/
int
bp_copyout(void *driverbuf, struct buf *bp, offset_t offset, size_t size)
{
return (bp_copy_common(BP_COPYOUT, bp, driverbuf, offset, size));
}
/*
* copy data from a buf_t which may not be mapped in, into a KVA.. offset
* is relative to the buf_t only.
*/
int
bp_copyin(struct buf *bp, void *driverbuf, offset_t offset, size_t size)
{
return (bp_copy_common(BP_COPYIN, bp, driverbuf, offset, size));
}
#define BP_COPY(dir, driverbuf, baddr, sz) \
(dir == BP_COPYIN) ? \
bcopy(baddr, driverbuf, sz) : bcopy(driverbuf, baddr, sz)
static int
bp_copy_common(bp_copydir_t dir, struct buf *bp, void *driverbuf,
offset_t offset, size_t size)
{
page_t **pplist;
uintptr_t poff;
uintptr_t voff;
struct as *as;
caddr_t kaddr;
caddr_t addr;
page_t *page;
size_t psize;
page_t *pp;
pfn_t pfn;
ASSERT((offset + size) <= bp->b_bcount);
/* if the buf_t already has a KVA, just do a bcopy */
if (!(bp->b_flags & (B_PHYS | B_PAGEIO))) {
BP_COPY(dir, driverbuf, bp->b_un.b_addr + offset, size);
return (0);
}
/* if we don't have kpm enabled, we need to do the slow path */
if (!kpm_enable || bp_force_copy) {
bp_mapin(bp);
BP_COPY(dir, driverbuf, bp->b_un.b_addr + offset, size);
bp_mapout(bp);
return (0);
}
/*
* kpm is enabled, and we need to map in the buf_t for the copy
*/
/* setup pp, plist, and make sure 'as' is right */
if (bp->b_flags & B_PAGEIO) {
pp = bp->b_pages;
pplist = NULL;
} else if (bp->b_flags & B_SHADOW) {
pp = NULL;
pplist = bp->b_shadow;
} else {
pp = NULL;
pplist = NULL;
if (bp->b_proc == NULL || (as = bp->b_proc->p_as) == NULL) {
as = &kas;
}
}
/*
* locals for the address, the offset into the first page, and the
* size of the first page we are going to copy.
*/
addr = (caddr_t)bp->b_un.b_addr;
poff = (uintptr_t)addr & PAGEOFFSET;
psize = MIN(PAGESIZE - poff, size);
/*
* we always start with a 0 offset into the driverbuf provided. The
* offset passed in only applies to the buf_t.
*/
voff = 0;
/* Loop until we've copied al the data */
while (size > 0) {
/*
* for a pp or pplist, get the pfn, then go to the next page_t
* for the next time around the loop.
*/
if (pp) {
page = pp;
pp = pp->p_next;
} else if (pplist != NULL) {
page = (*pplist);
pplist++;
/*
* We have a user VA. If we are going to copy this page, (e.g.
* the offset into the buf_t where we start to copy is
* within this page), get the pfn. Don't waste the cycles
* getting the pfn if we're not copying this page.
*/
} else if (offset < psize) {
pfn = hat_getpfnum(as->a_hat,
(caddr_t)((uintptr_t)addr & PAGEMASK));
if (pfn == PFN_INVALID) {
return (-1);
}
page = page_numtopp_nolock(pfn);
addr += psize - offset;
} else {
addr += psize;
}
/*
* if we have an initial offset into the buf_t passed in,
* and it falls within the current page, account for it in
* the page size (how much we will copy) and the offset into the
* page (where we'll start copying from).
*/
if ((offset > 0) && (offset < psize)) {
psize -= offset;
poff += offset;
offset = 0;
/*
* if we have an initial offset into the buf_t passed in,
* and it's not within the current page, skip this page.
* We don't have to worry about the first page offset and size
* anymore. psize will normally be PAGESIZE now unless we are
* on the last page.
*/
} else if (offset >= psize) {
offset -= psize;
psize = MIN(PAGESIZE, size);
poff = 0;
continue;
}
/*
* get a kpm mapping to the page, them copy in/out of the
* page. update size left and offset into the driverbuf passed
* in for the next time around the loop.
*/
kaddr = hat_kpm_mapin(page, NULL) + poff;
BP_COPY(dir, (void *)((uintptr_t)driverbuf + voff), kaddr,
psize);
hat_kpm_mapout(page, NULL, kaddr - poff);
size -= psize;
voff += psize;
poff = 0;
psize = MIN(PAGESIZE, size);
}
return (0);
}