UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity
OpenCloudOS-Kernel/drivers/char/agp/generic.c

1359 lines
35 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/*
* AGPGART driver.
* Copyright (C) 2004 Silicon Graphics, Inc.
* Copyright (C) 2002-2005 Dave Jones.
* Copyright (C) 1999 Jeff Hartmann.
* Copyright (C) 1999 Precision Insight, Inc.
* Copyright (C) 1999 Xi Graphics, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included
* in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* JEFF HARTMANN, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM,
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
* OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* TODO:
* - Allocate more than order 0 pages to avoid too much linear map splitting.
*/
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/miscdevice.h>
#include <linux/pm.h>
#include <linux/agp_backend.h>
#include <linux/vmalloc.h>
#include <linux/dma-mapping.h>
#include <linux/mm.h>
Detach sched.h from mm.h First thing mm.h does is including sched.h solely for can_do_mlock() inline function which has "current" dereference inside. By dealing with can_do_mlock() mm.h can be detached from sched.h which is good. See below, why. This patch a) removes unconditional inclusion of sched.h from mm.h b) makes can_do_mlock() normal function in mm/mlock.c c) exports can_do_mlock() to not break compilation d) adds sched.h inclusions back to files that were getting it indirectly. e) adds less bloated headers to some files (asm/signal.h, jiffies.h) that were getting them indirectly Net result is: a) mm.h users would get less code to open, read, preprocess, parse, ... if they don't need sched.h b) sched.h stops being dependency for significant number of files: on x86_64 allmodconfig touching sched.h results in recompile of 4083 files, after patch it's only 3744 (-8.3%). Cross-compile tested on all arm defconfigs, all mips defconfigs, all powerpc defconfigs, alpha alpha-up arm i386 i386-up i386-defconfig i386-allnoconfig ia64 ia64-up m68k mips parisc parisc-up powerpc powerpc-up s390 s390-up sparc sparc-up sparc64 sparc64-up um-x86_64 x86_64 x86_64-up x86_64-defconfig x86_64-allnoconfig as well as my two usual configs. Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-21 05:22:52 +08:00
#include <linux/sched.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
#include <asm/io.h>
#include <asm/cacheflush.h>
#include <asm/pgtable.h>
#include "agp.h"
__u32 *agp_gatt_table;
int agp_memory_reserved;
/*
* Needed by the Nforce GART driver for the time being. Would be
* nice to do this some other way instead of needing this export.
*/
EXPORT_SYMBOL_GPL(agp_memory_reserved);
/*
* Generic routines for handling agp_memory structures -
* They use the basic page allocation routines to do the brunt of the work.
*/
void agp_free_key(int key)
{
if (key < 0)
return;
if (key < MAXKEY)
clear_bit(key, agp_bridge->key_list);
}
EXPORT_SYMBOL(agp_free_key);
static int agp_get_key(void)
{
int bit;
bit = find_first_zero_bit(agp_bridge->key_list, MAXKEY);
if (bit < MAXKEY) {
set_bit(bit, agp_bridge->key_list);
return bit;
}
return -1;
}
agp: add chipset flushing support to AGP interface This bumps the AGP interface to 0.103. Certain Intel chipsets contains a global write buffer, and this can require flushing from the drm or X.org to make sure all data has hit RAM before initiating a GPU transfer, due to a lack of coherency with the integrated graphics device and this buffer. This just adds generic support to the AGP interfaces, a follow-on patch will add support to the Intel driver to use this interface. Signed-off-by: Dave Airlie <airlied@redhat.com>
2007-10-29 13:14:03 +08:00
void agp_flush_chipset(struct agp_bridge_data *bridge)
{
if (bridge->driver->chipset_flush)
bridge->driver->chipset_flush(bridge);
}
EXPORT_SYMBOL(agp_flush_chipset);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
/*
* Use kmalloc if possible for the page list. Otherwise fall back to
* vmalloc. This speeds things up and also saves memory for small AGP
* regions.
*/
void agp_alloc_page_array(size_t size, struct agp_memory *mem)
{
mem->memory = NULL;
mem->vmalloc_flag = 0;
[AGPGART] allow drm populated agp memory types cleanups Fix whitespace, braces, use kzalloc(). Cc: Dave Airlie <airlied@linux.ie> Cc: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Dave Jones <davej@redhat.com>
2007-02-06 08:09:35 +08:00
if (size <= 2*PAGE_SIZE)
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
mem->memory = kmalloc(size, GFP_KERNEL | __GFP_NORETRY);
if (mem->memory == NULL) {
mem->memory = vmalloc(size);
mem->vmalloc_flag = 1;
}
}
EXPORT_SYMBOL(agp_alloc_page_array);
void agp_free_page_array(struct agp_memory *mem)
{
if (mem->vmalloc_flag) {
vfree(mem->memory);
} else {
kfree(mem->memory);
}
}
EXPORT_SYMBOL(agp_free_page_array);
static struct agp_memory *agp_create_user_memory(unsigned long num_agp_pages)
{
struct agp_memory *new;
unsigned long alloc_size = num_agp_pages*sizeof(struct page *);
[AGPGART] allow drm populated agp memory types cleanups Fix whitespace, braces, use kzalloc(). Cc: Dave Airlie <airlied@linux.ie> Cc: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Dave Jones <davej@redhat.com>
2007-02-06 08:09:35 +08:00
new = kzalloc(sizeof(struct agp_memory), GFP_KERNEL);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
if (new == NULL)
return NULL;
new->key = agp_get_key();
if (new->key < 0) {
kfree(new);
return NULL;
}
agp_alloc_page_array(alloc_size, new);
if (new->memory == NULL) {
agp_free_key(new->key);
kfree(new);
return NULL;
}
new->num_scratch_pages = 0;
return new;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
struct agp_memory *agp_create_memory(int scratch_pages)
{
struct agp_memory *new;
[AGPGART] Replace kmalloc+memset's with kzalloc's Signed-off-by: Dave Jones <davej@redhat.com>
2005-10-21 06:12:16 +08:00
new = kzalloc(sizeof(struct agp_memory), GFP_KERNEL);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (new == NULL)
return NULL;
new->key = agp_get_key();
if (new->key < 0) {
kfree(new);
return NULL;
}
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
agp_alloc_page_array(PAGE_SIZE * scratch_pages, new);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (new->memory == NULL) {
agp_free_key(new->key);
kfree(new);
return NULL;
}
new->num_scratch_pages = scratch_pages;
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
new->type = AGP_NORMAL_MEMORY;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return new;
}
EXPORT_SYMBOL(agp_create_memory);
/**
* agp_free_memory - free memory associated with an agp_memory pointer.
*
* @curr: agp_memory pointer to be freed.
*
* It is the only function that can be called when the backend is not owned
* by the caller. (So it can free memory on client death.)
*/
void agp_free_memory(struct agp_memory *curr)
{
size_t i;
if (curr == NULL)
return;
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
if (curr->is_bound)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
agp_unbind_memory(curr);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
if (curr->type >= AGP_USER_TYPES) {
agp_generic_free_by_type(curr);
return;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (curr->type != 0) {
curr->bridge->driver->free_by_type(curr);
return;
}
if (curr->page_count != 0) {
for (i = 0; i < curr->page_count; i++) {
agp: two-stage page destruction issue besides it apparently being useful only in 2.6.24 (the changes in 2.6.25 really mean that it could be converted back to a single-stage mechanism), I'm seeing an issue in Xen Dom0 kernels, which is caused by the calling of gart_to_virt() in the second stage invocations of the destroy function. I think that besides this being a real issue with Xen (where unmap_page_from_agp() is not just a page table attribute change), this also is invalid from a theoretical perspective: One should not assume that gart_to_virt() is still valid after unmapping a page. So minimally (keeping the 2-stage mechanism) a patch like the one below would be needed. Jan Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-06-18 16:28:00 +08:00
curr->memory[i] = (unsigned long)gart_to_virt(curr->memory[i]);
curr->bridge->driver->agp_destroy_page((void *)curr->memory[i],
AGP_PAGE_DESTROY_UNMAP);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
AGP fix race condition between unmapping and freeing pages With Andi's clflush fixup, we were getting hangs on server exit, flushing the mappings after freeing each page helped. This showed up a race condition where the pages after being freed could be reused before the agp mappings had been flushed. Flushing after each single page is a bad thing for future drm work, so make the page destroy a two pass unmapping all the pages, flushing the mappings, and then destroying the pages. Signed-off-by: Dave Airlie <airlied@linux.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-10-15 08:19:16 +08:00
for (i = 0; i < curr->page_count; i++) {
agp: two-stage page destruction issue besides it apparently being useful only in 2.6.24 (the changes in 2.6.25 really mean that it could be converted back to a single-stage mechanism), I'm seeing an issue in Xen Dom0 kernels, which is caused by the calling of gart_to_virt() in the second stage invocations of the destroy function. I think that besides this being a real issue with Xen (where unmap_page_from_agp() is not just a page table attribute change), this also is invalid from a theoretical perspective: One should not assume that gart_to_virt() is still valid after unmapping a page. So minimally (keeping the 2-stage mechanism) a patch like the one below would be needed. Jan Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-06-18 16:28:00 +08:00
curr->bridge->driver->agp_destroy_page((void *)curr->memory[i],
AGP_PAGE_DESTROY_FREE);
AGP fix race condition between unmapping and freeing pages With Andi's clflush fixup, we were getting hangs on server exit, flushing the mappings after freeing each page helped. This showed up a race condition where the pages after being freed could be reused before the agp mappings had been flushed. Flushing after each single page is a bad thing for future drm work, so make the page destroy a two pass unmapping all the pages, flushing the mappings, and then destroying the pages. Signed-off-by: Dave Airlie <airlied@linux.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-10-15 08:19:16 +08:00
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
agp_free_key(curr->key);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
agp_free_page_array(curr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
kfree(curr);
}
EXPORT_SYMBOL(agp_free_memory);
#define ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long))
/**
* agp_allocate_memory - allocate a group of pages of a certain type.
*
* @page_count: size_t argument of the number of pages
* @type: u32 argument of the type of memory to be allocated.
*
* Every agp bridge device will allow you to allocate AGP_NORMAL_MEMORY which
* maps to physical ram. Any other type is device dependent.
*
* It returns NULL whenever memory is unavailable.
*/
struct agp_memory *agp_allocate_memory(struct agp_bridge_data *bridge,
size_t page_count, u32 type)
{
int scratch_pages;
struct agp_memory *new;
size_t i;
if (!bridge)
return NULL;
if ((atomic_read(&bridge->current_memory_agp) + page_count) > bridge->max_memory_agp)
return NULL;
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
if (type >= AGP_USER_TYPES) {
new = agp_generic_alloc_user(page_count, type);
if (new)
new->bridge = bridge;
return new;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (type != 0) {
new = bridge->driver->alloc_by_type(page_count, type);
if (new)
new->bridge = bridge;
return new;
}
scratch_pages = (page_count + ENTRIES_PER_PAGE - 1) / ENTRIES_PER_PAGE;
new = agp_create_memory(scratch_pages);
if (new == NULL)
return NULL;
for (i = 0; i < page_count; i++) {
void *addr = bridge->driver->agp_alloc_page(bridge);
if (addr == NULL) {
agp_free_memory(new);
return NULL;
}
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
new->memory[i] = virt_to_gart(addr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
new->page_count++;
}
[PATCH] AGP performance fixes AGP allocation/deallocation is suffering major performance issues due to the nature of global_flush_tlb() being called on every change_page_attr() call. For small allocations this isn't really seen, but when you start allocating 50000 pages of AGP space, for say, texture memory, then things can take seconds to complete. In some cases the situation is doubled or even quadrupled in the time due to SMP, or a deallocation, then a new reallocation. I've had a case of upto 20 seconds wait time to deallocate and reallocate AGP space. This patch fixes the problem by making it the caller's responsibility to call global_flush_tlb(), and so removes it from every instance of mapping a page into AGP space until the time that all change_page_attr() changes are done. Signed-off-by: Dave Jones <davej@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org>
2005-11-07 15:35:34 +08:00
new->bridge = bridge;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return new;
}
EXPORT_SYMBOL(agp_allocate_memory);
/* End - Generic routines for handling agp_memory structures */
static int agp_return_size(void)
{
int current_size;
void *temp;
temp = agp_bridge->current_size;
switch (agp_bridge->driver->size_type) {
case U8_APER_SIZE:
current_size = A_SIZE_8(temp)->size;
break;
case U16_APER_SIZE:
current_size = A_SIZE_16(temp)->size;
break;
case U32_APER_SIZE:
current_size = A_SIZE_32(temp)->size;
break;
case LVL2_APER_SIZE:
current_size = A_SIZE_LVL2(temp)->size;
break;
case FIXED_APER_SIZE:
current_size = A_SIZE_FIX(temp)->size;
break;
default:
current_size = 0;
break;
}
current_size -= (agp_memory_reserved / (1024*1024));
if (current_size <0)
current_size = 0;
return current_size;
}
int agp_num_entries(void)
{
int num_entries;
void *temp;
temp = agp_bridge->current_size;
switch (agp_bridge->driver->size_type) {
case U8_APER_SIZE:
num_entries = A_SIZE_8(temp)->num_entries;
break;
case U16_APER_SIZE:
num_entries = A_SIZE_16(temp)->num_entries;
break;
case U32_APER_SIZE:
num_entries = A_SIZE_32(temp)->num_entries;
break;
case LVL2_APER_SIZE:
num_entries = A_SIZE_LVL2(temp)->num_entries;
break;
case FIXED_APER_SIZE:
num_entries = A_SIZE_FIX(temp)->num_entries;
break;
default:
num_entries = 0;
break;
}
num_entries -= agp_memory_reserved>>PAGE_SHIFT;
if (num_entries<0)
num_entries = 0;
return num_entries;
}
EXPORT_SYMBOL_GPL(agp_num_entries);
/**
* agp_copy_info - copy bridge state information
*
[AGPGART] Lots of CodingStyle/whitespace cleanups. Eliminate trailing whitespace. s/if(/if (/ s/for(/for (/ Signed-off-by: Dave Jones <davej@redhat.com>
2006-02-28 13:54:25 +08:00
* @info: agp_kern_info pointer. The caller should insure that this pointer is valid.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*
* This function copies information about the agp bridge device and the state of
* the agp backend into an agp_kern_info pointer.
*/
int agp_copy_info(struct agp_bridge_data *bridge, struct agp_kern_info *info)
{
memset(info, 0, sizeof(struct agp_kern_info));
if (!bridge) {
info->chipset = NOT_SUPPORTED;
return -EIO;
}
info->version.major = bridge->version->major;
info->version.minor = bridge->version->minor;
info->chipset = SUPPORTED;
info->device = bridge->dev;
[PATCH] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). [AGPGART] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). As mentioned earlier, the current check_bridge_mode() code assumes that AGP bridges are PCI devices. This isn't always true. Definitely not for HP zx1 chipset and the same seems to be the case for SGI's AGP bridge. The patch below fixes the problem by picking up the AGP_MODE_3_0 bit from bridge->mode. I feel like I may be missing something, since I can't see any reason why check_bridge_mode() wasn't doing that in the first place. According to the AGP 3.0 specs, the AGP_MODE_3_0 bit is determined during the hardware reset and cannot be changed, so it seems to me it should be safe to pick it up from bridge->mode. With the patch applied, I can definitely use AGP acceleration both with AGP 2.0 and AGP 3.0 (one with an Nvidia card, the other with an ATI FireGL card). Unless someone spots a problem, please apply this patch so 3d acceleration can work on zx1 boxes again. This makes AGP work again on machines with an AGP bridge that isn't a PCI device. Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com> Signed-off-by: Dave Jones <davej@redhat.com>
2005-04-05 04:29:43 +08:00
if (bridge->mode & AGPSTAT_MODE_3_0)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
info->mode = bridge->mode & ~AGP3_RESERVED_MASK;
else
info->mode = bridge->mode & ~AGP2_RESERVED_MASK;
info->aper_base = bridge->gart_bus_addr;
info->aper_size = agp_return_size();
info->max_memory = bridge->max_memory_agp;
info->current_memory = atomic_read(&bridge->current_memory_agp);
info->cant_use_aperture = bridge->driver->cant_use_aperture;
info->vm_ops = bridge->vm_ops;
info->page_mask = ~0UL;
return 0;
}
EXPORT_SYMBOL(agp_copy_info);
/* End - Routine to copy over information structure */
/*
* Routines for handling swapping of agp_memory into the GATT -
* These routines take agp_memory and insert them into the GATT.
* They call device specific routines to actually write to the GATT.
*/
/**
* agp_bind_memory - Bind an agp_memory structure into the GATT.
*
* @curr: agp_memory pointer
* @pg_start: an offset into the graphics aperture translation table
*
* It returns -EINVAL if the pointer == NULL.
* It returns -EBUSY if the area of the table requested is already in use.
*/
int agp_bind_memory(struct agp_memory *curr, off_t pg_start)
{
int ret_val;
if (curr == NULL)
return -EINVAL;
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
if (curr->is_bound) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "memory %p is already bound!\n", curr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return -EINVAL;
}
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
if (!curr->is_flushed) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
curr->bridge->driver->cache_flush();
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
curr->is_flushed = true;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
ret_val = curr->bridge->driver->insert_memory(curr, pg_start, curr->type);
if (ret_val != 0)
return ret_val;
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
curr->is_bound = true;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
curr->pg_start = pg_start;
return 0;
}
EXPORT_SYMBOL(agp_bind_memory);
/**
* agp_unbind_memory - Removes an agp_memory structure from the GATT
*
* @curr: agp_memory pointer to be removed from the GATT.
*
* It returns -EINVAL if this piece of agp_memory is not currently bound to
* the graphics aperture translation table or if the agp_memory pointer == NULL
*/
int agp_unbind_memory(struct agp_memory *curr)
{
int ret_val;
if (curr == NULL)
return -EINVAL;
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
if (!curr->is_bound) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "memory %p was not bound!\n", curr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return -EINVAL;
}
ret_val = curr->bridge->driver->remove_memory(curr, curr->pg_start, curr->type);
if (ret_val != 0)
return ret_val;
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
curr->is_bound = false;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
curr->pg_start = 0;
return 0;
}
EXPORT_SYMBOL(agp_unbind_memory);
/* End - Routines for handling swapping of agp_memory into the GATT */
/* Generic Agp routines - Start */
static void agp_v2_parse_one(u32 *requested_mode, u32 *bridge_agpstat, u32 *vga_agpstat)
{
u32 tmp;
if (*requested_mode & AGP2_RESERVED_MASK) {
[AGPGART] When we encounter reserved mode bits, print them out. Signed-off-by: Dave Jones <davej@redhat.com>
2005-11-05 07:18:56 +08:00
printk(KERN_INFO PFX "reserved bits set (%x) in mode 0x%x. Fixed.\n",
*requested_mode & AGP2_RESERVED_MASK, *requested_mode);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*requested_mode &= ~AGP2_RESERVED_MASK;
}
[AGPGART] Fix up misprogrammed bridges with incorrect AGPv2 rates. Some dumb bridges are programmed to disobey the AGP2 spec. This is likely a BIOS misprogramming rather than poweron default, or it would be a lot more common. AGPv2 spec 6.1.9 states: "The RATE field indicates the data transfer rates supported by this device. A.G.P. devices must report all that apply." Fix them up as best we can. This will prevent errors like.. agpgart: Found an AGP 3.5 compliant device at 0000:00:00.0. agpgart: req mode 1f000201 bridge_agpstat 1f000a14 vga_agpstat 2f000217. agpgart: Device is in legacy mode, falling back to 2.x agpgart: Putting AGP V2 device at 0000:00:00.0 into 0x mode agpgart: Putting AGP V2 device at 0000:01:00.0 into 0x mode agpgart: Putting AGP V2 device at 0000:01:00.1 into 0x mode https://bugs.freedesktop.org/show_bug.cgi?id=8816 Signed-off-by: Dave Jones <davej@redhat.com>
2006-11-04 04:13:27 +08:00
/*
* Some dumb bridges are programmed to disobey the AGP2 spec.
* This is likely a BIOS misprogramming rather than poweron default, or
* it would be a lot more common.
* https://bugs.freedesktop.org/show_bug.cgi?id=8816
* AGPv2 spec 6.1.9 states:
* The RATE field indicates the data transfer rates supported by this
* device. A.G.P. devices must report all that apply.
* Fix them up as best we can.
*/
switch (*bridge_agpstat & 7) {
case 4:
*bridge_agpstat |= (AGPSTAT2_2X | AGPSTAT2_1X);
printk(KERN_INFO PFX "BIOS bug. AGP bridge claims to only support x4 rate"
"Fixing up support for x2 & x1\n");
break;
case 2:
*bridge_agpstat |= AGPSTAT2_1X;
printk(KERN_INFO PFX "BIOS bug. AGP bridge claims to only support x2 rate"
"Fixing up support for x1\n");
break;
default:
break;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/* Check the speed bits make sense. Only one should be set. */
tmp = *requested_mode & 7;
switch (tmp) {
case 0:
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s tried to set rate=x0. Setting to x1 mode.\n", current->comm);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*requested_mode |= AGPSTAT2_1X;
break;
case 1:
case 2:
break;
case 3:
*requested_mode &= ~(AGPSTAT2_1X); /* rate=2 */
break;
case 4:
break;
case 5:
case 6:
case 7:
*requested_mode &= ~(AGPSTAT2_1X|AGPSTAT2_2X); /* rate=4*/
break;
}
/* disable SBA if it's not supported */
if (!((*bridge_agpstat & AGPSTAT_SBA) && (*vga_agpstat & AGPSTAT_SBA) && (*requested_mode & AGPSTAT_SBA)))
*bridge_agpstat &= ~AGPSTAT_SBA;
/* Set rate */
if (!((*bridge_agpstat & AGPSTAT2_4X) && (*vga_agpstat & AGPSTAT2_4X) && (*requested_mode & AGPSTAT2_4X)))
*bridge_agpstat &= ~AGPSTAT2_4X;
if (!((*bridge_agpstat & AGPSTAT2_2X) && (*vga_agpstat & AGPSTAT2_2X) && (*requested_mode & AGPSTAT2_2X)))
*bridge_agpstat &= ~AGPSTAT2_2X;
if (!((*bridge_agpstat & AGPSTAT2_1X) && (*vga_agpstat & AGPSTAT2_1X) && (*requested_mode & AGPSTAT2_1X)))
*bridge_agpstat &= ~AGPSTAT2_1X;
/* Now we know what mode it should be, clear out the unwanted bits. */
if (*bridge_agpstat & AGPSTAT2_4X)
*bridge_agpstat &= ~(AGPSTAT2_1X | AGPSTAT2_2X); /* 4X */
if (*bridge_agpstat & AGPSTAT2_2X)
*bridge_agpstat &= ~(AGPSTAT2_1X | AGPSTAT2_4X); /* 2X */
if (*bridge_agpstat & AGPSTAT2_1X)
*bridge_agpstat &= ~(AGPSTAT2_2X | AGPSTAT2_4X); /* 1X */
/* Apply any errata. */
if (agp_bridge->flags & AGP_ERRATA_FASTWRITES)
*bridge_agpstat &= ~AGPSTAT_FW;
if (agp_bridge->flags & AGP_ERRATA_SBA)
*bridge_agpstat &= ~AGPSTAT_SBA;
if (agp_bridge->flags & AGP_ERRATA_1X) {
*bridge_agpstat &= ~(AGPSTAT2_2X | AGPSTAT2_4X);
*bridge_agpstat |= AGPSTAT2_1X;
}
/* If we've dropped down to 1X, disable fast writes. */
if (*bridge_agpstat & AGPSTAT2_1X)
*bridge_agpstat &= ~AGPSTAT_FW;
}
/*
* requested_mode = Mode requested by (typically) X.
* bridge_agpstat = PCI_AGP_STATUS from agp bridge.
* vga_agpstat = PCI_AGP_STATUS from graphic card.
*/
static void agp_v3_parse_one(u32 *requested_mode, u32 *bridge_agpstat, u32 *vga_agpstat)
{
u32 origbridge=*bridge_agpstat, origvga=*vga_agpstat;
u32 tmp;
if (*requested_mode & AGP3_RESERVED_MASK) {
[AGPGART] When we encounter reserved mode bits, print them out. Signed-off-by: Dave Jones <davej@redhat.com>
2005-11-05 07:18:56 +08:00
printk(KERN_INFO PFX "reserved bits set (%x) in mode 0x%x. Fixed.\n",
*requested_mode & AGP3_RESERVED_MASK, *requested_mode);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*requested_mode &= ~AGP3_RESERVED_MASK;
}
/* Check the speed bits make sense. */
tmp = *requested_mode & 7;
if (tmp == 0) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s tried to set rate=x0. Setting to AGP3 x4 mode.\n", current->comm);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*requested_mode |= AGPSTAT3_4X;
}
if (tmp >= 3) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s tried to set rate=x%d. Setting to AGP3 x8 mode.\n", current->comm, tmp * 4);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*requested_mode = (*requested_mode & ~7) | AGPSTAT3_8X;
}
/* ARQSZ - Set the value to the maximum one.
* Don't allow the mode register to override values. */
*bridge_agpstat = ((*bridge_agpstat & ~AGPSTAT_ARQSZ) |
max_t(u32,(*bridge_agpstat & AGPSTAT_ARQSZ),(*vga_agpstat & AGPSTAT_ARQSZ)));
/* Calibration cycle.
* Don't allow the mode register to override values. */
*bridge_agpstat = ((*bridge_agpstat & ~AGPSTAT_CAL_MASK) |
min_t(u32,(*bridge_agpstat & AGPSTAT_CAL_MASK),(*vga_agpstat & AGPSTAT_CAL_MASK)));
/* SBA *must* be supported for AGP v3 */
*bridge_agpstat |= AGPSTAT_SBA;
/*
* Set speed.
* Check for invalid speeds. This can happen when applications
* written before the AGP 3.0 standard pass AGP2.x modes to AGP3 hardware
*/
if (*requested_mode & AGPSTAT_MODE_3_0) {
/*
* Caller hasn't a clue what it is doing. Bridge is in 3.0 mode,
* have been passed a 3.0 mode, but with 2.x speed bits set.
* AGP2.x 4x -> AGP3.0 4x.
*/
if (*requested_mode & AGPSTAT2_4X) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s passes broken AGP3 flags (%x). Fixed.\n",
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
current->comm, *requested_mode);
*requested_mode &= ~AGPSTAT2_4X;
*requested_mode |= AGPSTAT3_4X;
}
} else {
/*
* The caller doesn't know what they are doing. We are in 3.0 mode,
* but have been passed an AGP 2.x mode.
* Convert AGP 1x,2x,4x -> AGP 3.0 4x.
*/
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s passes broken AGP2 flags (%x) in AGP3 mode. Fixed.\n",
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
current->comm, *requested_mode);
*requested_mode &= ~(AGPSTAT2_4X | AGPSTAT2_2X | AGPSTAT2_1X);
*requested_mode |= AGPSTAT3_4X;
}
if (*requested_mode & AGPSTAT3_8X) {
if (!(*bridge_agpstat & AGPSTAT3_8X)) {
*bridge_agpstat &= ~(AGPSTAT3_8X | AGPSTAT3_RSVD);
*bridge_agpstat |= AGPSTAT3_4X;
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s requested AGPx8 but bridge not capable.\n", current->comm);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return;
}
if (!(*vga_agpstat & AGPSTAT3_8X)) {
*bridge_agpstat &= ~(AGPSTAT3_8X | AGPSTAT3_RSVD);
*bridge_agpstat |= AGPSTAT3_4X;
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "%s requested AGPx8 but graphic card not capable.\n", current->comm);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return;
}
/* All set, bridge & device can do AGP x8*/
*bridge_agpstat &= ~(AGPSTAT3_4X | AGPSTAT3_RSVD);
goto done;
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
} else if (*requested_mode & AGPSTAT3_4X) {
*bridge_agpstat &= ~(AGPSTAT3_8X | AGPSTAT3_RSVD);
*bridge_agpstat |= AGPSTAT3_4X;
goto done;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
} else {
/*
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
* If we didn't specify an AGP mode, we see if both
* the graphics card, and the bridge can do x8, and use if so.
* If not, we fall back to x4 mode.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
*/
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
if ((*bridge_agpstat & AGPSTAT3_8X) && (*vga_agpstat & AGPSTAT3_8X)) {
[AGPGART] printk fixups. Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-29 07:50:07 +08:00
printk(KERN_INFO PFX "No AGP mode specified. Setting to highest mode "
"supported by bridge & card (x8).\n");
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
*bridge_agpstat &= ~(AGPSTAT3_4X | AGPSTAT3_RSVD);
*vga_agpstat &= ~(AGPSTAT3_4X | AGPSTAT3_RSVD);
} else {
printk(KERN_INFO PFX "Fell back to AGPx4 mode because");
if (!(*bridge_agpstat & AGPSTAT3_8X)) {
[AGPGART] printk fixups. Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-29 07:50:07 +08:00
printk(KERN_INFO PFX "bridge couldn't do x8. bridge_agpstat:%x (orig=%x)\n",
*bridge_agpstat, origbridge);
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
*bridge_agpstat &= ~(AGPSTAT3_8X | AGPSTAT3_RSVD);
*bridge_agpstat |= AGPSTAT3_4X;
}
if (!(*vga_agpstat & AGPSTAT3_8X)) {
[AGPGART] printk fixups. Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-29 07:50:07 +08:00
printk(KERN_INFO PFX "graphics card couldn't do x8. vga_agpstat:%x (orig=%x)\n",
*vga_agpstat, origvga);
[AGPGART] Rework AGPv3 modesetting fallback. Sometimes the logic to handle AGPx8->AGPx4 fallback failed, as can be seen in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=197346 The failures occured if the bridge was in AGPx8 mode, but the user hadn't specified a mode in their X config. We weren't setting the mode to the highest mode capable by the video card+bridge (as we do in the AGPv2 case), which was leading to all kinds of mayhem including us believing that after falling back from AGPx8, that we couldn't do x4 mode (which is disastrous in AGPv3, as those are the only two modes possible). Signed-off-by: Dave Jones <davej@redhat.com>
2006-09-11 09:12:20 +08:00
*vga_agpstat &= ~(AGPSTAT3_8X | AGPSTAT3_RSVD);
*vga_agpstat |= AGPSTAT3_4X;
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
}
done:
/* Apply any errata. */
if (agp_bridge->flags & AGP_ERRATA_FASTWRITES)
*bridge_agpstat &= ~AGPSTAT_FW;
if (agp_bridge->flags & AGP_ERRATA_SBA)
*bridge_agpstat &= ~AGPSTAT_SBA;
if (agp_bridge->flags & AGP_ERRATA_1X) {
*bridge_agpstat &= ~(AGPSTAT2_2X | AGPSTAT2_4X);
*bridge_agpstat |= AGPSTAT2_1X;
}
}
/**
* agp_collect_device_status - determine correct agp_cmd from various agp_stat's
* @bridge: an agp_bridge_data struct allocated for the AGP host bridge.
* @requested_mode: requested agp_stat from userspace (Typically from X)
* @bridge_agpstat: current agp_stat from AGP bridge.
*
* This function will hunt for an AGP graphics card, and try to match
* the requested mode to the capabilities of both the bridge and the card.
*/
u32 agp_collect_device_status(struct agp_bridge_data *bridge, u32 requested_mode, u32 bridge_agpstat)
{
struct pci_dev *device = NULL;
u32 vga_agpstat;
u8 cap_ptr;
for (;;) {
device = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, device);
if (!device) {
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "Couldn't find an AGP VGA controller.\n");
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return 0;
}
cap_ptr = pci_find_capability(device, PCI_CAP_ID_AGP);
if (cap_ptr)
break;
}
/*
* Ok, here we have a AGP device. Disable impossible
* settings, and adjust the readqueue to the minimum.
*/
pci_read_config_dword(device, cap_ptr+PCI_AGP_STATUS, &vga_agpstat);
/* adjust RQ depth */
bridge_agpstat = ((bridge_agpstat & ~AGPSTAT_RQ_DEPTH) |
min_t(u32, (requested_mode & AGPSTAT_RQ_DEPTH),
min_t(u32, (bridge_agpstat & AGPSTAT_RQ_DEPTH), (vga_agpstat & AGPSTAT_RQ_DEPTH))));
/* disable FW if it's not supported */
if (!((bridge_agpstat & AGPSTAT_FW) &&
(vga_agpstat & AGPSTAT_FW) &&
(requested_mode & AGPSTAT_FW)))
bridge_agpstat &= ~AGPSTAT_FW;
/* Check to see if we are operating in 3.0 mode */
[PATCH] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). [AGPGART] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). As mentioned earlier, the current check_bridge_mode() code assumes that AGP bridges are PCI devices. This isn't always true. Definitely not for HP zx1 chipset and the same seems to be the case for SGI's AGP bridge. The patch below fixes the problem by picking up the AGP_MODE_3_0 bit from bridge->mode. I feel like I may be missing something, since I can't see any reason why check_bridge_mode() wasn't doing that in the first place. According to the AGP 3.0 specs, the AGP_MODE_3_0 bit is determined during the hardware reset and cannot be changed, so it seems to me it should be safe to pick it up from bridge->mode. With the patch applied, I can definitely use AGP acceleration both with AGP 2.0 and AGP 3.0 (one with an Nvidia card, the other with an ATI FireGL card). Unless someone spots a problem, please apply this patch so 3d acceleration can work on zx1 boxes again. This makes AGP work again on machines with an AGP bridge that isn't a PCI device. Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com> Signed-off-by: Dave Jones <davej@redhat.com>
2005-04-05 04:29:43 +08:00
if (agp_bridge->mode & AGPSTAT_MODE_3_0)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
agp_v3_parse_one(&requested_mode, &bridge_agpstat, &vga_agpstat);
else
agp_v2_parse_one(&requested_mode, &bridge_agpstat, &vga_agpstat);
pci_dev_put(device);
return bridge_agpstat;
}
EXPORT_SYMBOL(agp_collect_device_status);
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
void agp_device_command(u32 bridge_agpstat, bool agp_v3)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
{
struct pci_dev *device = NULL;
int mode;
mode = bridge_agpstat & 0x7;
if (agp_v3)
mode *= 4;
for_each_pci_dev(device) {
u8 agp = pci_find_capability(device, PCI_CAP_ID_AGP);
if (!agp)
continue;
printk(KERN_INFO PFX "Putting AGP V%d device at %s into %dx mode\n",
agp_v3 ? 3 : 2, pci_name(device), mode);
pci_write_config_dword(device, agp + PCI_AGP_COMMAND, bridge_agpstat);
}
}
EXPORT_SYMBOL(agp_device_command);
void get_agp_version(struct agp_bridge_data *bridge)
{
u32 ncapid;
/* Exit early if already set by errata workarounds. */
if (bridge->major_version != 0)
return;
pci_read_config_dword(bridge->dev, bridge->capndx, &ncapid);
bridge->major_version = (ncapid >> AGP_MAJOR_VERSION_SHIFT) & 0xf;
bridge->minor_version = (ncapid >> AGP_MINOR_VERSION_SHIFT) & 0xf;
}
EXPORT_SYMBOL(get_agp_version);
void agp_generic_enable(struct agp_bridge_data *bridge, u32 requested_mode)
{
u32 bridge_agpstat, temp;
get_agp_version(agp_bridge);
printk(KERN_INFO PFX "Found an AGP %d.%d compliant device at %s.\n",
agp_bridge->major_version,
agp_bridge->minor_version,
pci_name(agp_bridge->dev));
pci_read_config_dword(agp_bridge->dev,
agp_bridge->capndx + PCI_AGP_STATUS, &bridge_agpstat);
bridge_agpstat = agp_collect_device_status(agp_bridge, requested_mode, bridge_agpstat);
if (bridge_agpstat == 0)
/* Something bad happened. FIXME: Return error code? */
return;
bridge_agpstat |= AGPSTAT_AGP_ENABLE;
/* Do AGP version specific frobbing. */
if (bridge->major_version >= 3) {
[PATCH] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). [AGPGART] Replace check_bridge_mode() with (bridge->mode & AGSTAT_MODE_3_0). As mentioned earlier, the current check_bridge_mode() code assumes that AGP bridges are PCI devices. This isn't always true. Definitely not for HP zx1 chipset and the same seems to be the case for SGI's AGP bridge. The patch below fixes the problem by picking up the AGP_MODE_3_0 bit from bridge->mode. I feel like I may be missing something, since I can't see any reason why check_bridge_mode() wasn't doing that in the first place. According to the AGP 3.0 specs, the AGP_MODE_3_0 bit is determined during the hardware reset and cannot be changed, so it seems to me it should be safe to pick it up from bridge->mode. With the patch applied, I can definitely use AGP acceleration both with AGP 2.0 and AGP 3.0 (one with an Nvidia card, the other with an ATI FireGL card). Unless someone spots a problem, please apply this patch so 3d acceleration can work on zx1 boxes again. This makes AGP work again on machines with an AGP bridge that isn't a PCI device. Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com> Signed-off-by: Dave Jones <davej@redhat.com>
2005-04-05 04:29:43 +08:00
if (bridge->mode & AGPSTAT_MODE_3_0) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/* If we have 3.5, we can do the isoch stuff. */
if (bridge->minor_version >= 5)
agp_3_5_enable(bridge);
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
agp_device_command(bridge_agpstat, true);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return;
} else {
/* Disable calibration cycle in RX91<1> when not in AGP3.0 mode of operation.*/
bridge_agpstat &= ~(7<<10) ;
pci_read_config_dword(bridge->dev,
bridge->capndx+AGPCTRL, &temp);
temp |= (1<<9);
pci_write_config_dword(bridge->dev,
bridge->capndx+AGPCTRL, temp);
Fix up various printk levels and whitespace corrections. Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-08-18 14:08:11 +08:00
printk(KERN_INFO PFX "Device is in legacy mode,"
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
" falling back to 2.x\n");
}
}
/* AGP v<3 */
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
agp_device_command(bridge_agpstat, false);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
EXPORT_SYMBOL(agp_generic_enable);
int agp_generic_create_gatt_table(struct agp_bridge_data *bridge)
{
char *table;
char *table_end;
int size;
int page_order;
int num_entries;
int i;
void *temp;
struct page *page;
/* The generic routines can't handle 2 level gatt's */
if (bridge->driver->size_type == LVL2_APER_SIZE)
return -EINVAL;
table = NULL;
i = bridge->aperture_size_idx;
temp = bridge->current_size;
size = page_order = num_entries = 0;
if (bridge->driver->size_type != FIXED_APER_SIZE) {
do {
switch (bridge->driver->size_type) {
case U8_APER_SIZE:
size = A_SIZE_8(temp)->size;
page_order =
A_SIZE_8(temp)->page_order;
num_entries =
A_SIZE_8(temp)->num_entries;
break;
case U16_APER_SIZE:
size = A_SIZE_16(temp)->size;
page_order = A_SIZE_16(temp)->page_order;
num_entries = A_SIZE_16(temp)->num_entries;
break;
case U32_APER_SIZE:
size = A_SIZE_32(temp)->size;
page_order = A_SIZE_32(temp)->page_order;
num_entries = A_SIZE_32(temp)->num_entries;
break;
/* This case will never really happen. */
case FIXED_APER_SIZE:
case LVL2_APER_SIZE:
default:
size = page_order = num_entries = 0;
break;
}
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
table = alloc_gatt_pages(page_order);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (table == NULL) {
i++;
switch (bridge->driver->size_type) {
case U8_APER_SIZE:
bridge->current_size = A_IDX8(bridge);
break;
case U16_APER_SIZE:
bridge->current_size = A_IDX16(bridge);
break;
case U32_APER_SIZE:
bridge->current_size = A_IDX32(bridge);
break;
[AGPGART] Remove pointless code from agp_generic_create_gatt_table() Signed-off-by: Dave Jones <davej@redhat.com>
2006-05-31 06:19:39 +08:00
/* These cases will never really happen. */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
case FIXED_APER_SIZE:
case LVL2_APER_SIZE:
default:
break;
}
temp = bridge->current_size;
} else {
bridge->aperture_size_idx = i;
}
} while (!table && (i < bridge->driver->num_aperture_sizes));
} else {
size = ((struct aper_size_info_fixed *) temp)->size;
page_order = ((struct aper_size_info_fixed *) temp)->page_order;
num_entries = ((struct aper_size_info_fixed *) temp)->num_entries;
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
table = alloc_gatt_pages(page_order);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
if (table == NULL)
return -ENOMEM;
table_end = table + ((PAGE_SIZE * (1 << page_order)) - 1);
for (page = virt_to_page(table); page <= virt_to_page(table_end); page++)
SetPageReserved(page);
bridge->gatt_table_real = (u32 *) table;
agp_gatt_table = (void *)table;
bridge->driver->cache_flush();
fix historic ioremap() abuse in AGP Several AGP drivers right now use ioremap_nocache() on kernel ram in order to turn a page of regular memory uncached. There are two problems with this: 1) This is a total nightmare for the ioremap() implementation to keep various mappings of the same page coherent. 2) It's a total nightmare for the AGP code since it adds a ton of complexity in terms of keeping track of 2 different pointers to the same thing, in terms of error handling etc etc. This patch fixes this by making the AGP drivers use the new set_memory_XX APIs instead. Note: amd-k7-agp.c is built on Alpha too, and generic.c is built on ia64 as well, which do not yet have the set_memory_*() APIs, so for them some we have a few ugly #ifdefs - hopefully they'll be fixed soon. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Dave Airlie <airlied@linux.ie>
2008-02-06 12:16:00 +08:00
#ifdef CONFIG_X86
set_memory_uc((unsigned long)table, 1 << page_order);
bridge->gatt_table = (void *)table;
#else
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
bridge->gatt_table = ioremap_nocache(virt_to_gart(table),
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
(PAGE_SIZE * (1 << page_order)));
bridge->driver->cache_flush();
fix historic ioremap() abuse in AGP Several AGP drivers right now use ioremap_nocache() on kernel ram in order to turn a page of regular memory uncached. There are two problems with this: 1) This is a total nightmare for the ioremap() implementation to keep various mappings of the same page coherent. 2) It's a total nightmare for the AGP code since it adds a ton of complexity in terms of keeping track of 2 different pointers to the same thing, in terms of error handling etc etc. This patch fixes this by making the AGP drivers use the new set_memory_XX APIs instead. Note: amd-k7-agp.c is built on Alpha too, and generic.c is built on ia64 as well, which do not yet have the set_memory_*() APIs, so for them some we have a few ugly #ifdefs - hopefully they'll be fixed soon. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Dave Airlie <airlied@linux.ie>
2008-02-06 12:16:00 +08:00
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (bridge->gatt_table == NULL) {
for (page = virt_to_page(table); page <= virt_to_page(table_end); page++)
ClearPageReserved(page);
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
free_gatt_pages(table, page_order);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
return -ENOMEM;
}
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
bridge->gatt_bus_addr = virt_to_gart(bridge->gatt_table_real);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/* AK: bogus, should encode addresses > 4GB */
for (i = 0; i < num_entries; i++) {
writel(bridge->scratch_page, bridge->gatt_table+i);
readl(bridge->gatt_table+i); /* PCI Posting. */
}
return 0;
}
EXPORT_SYMBOL(agp_generic_create_gatt_table);
int agp_generic_free_gatt_table(struct agp_bridge_data *bridge)
{
int page_order;
char *table, *table_end;
void *temp;
struct page *page;
temp = bridge->current_size;
switch (bridge->driver->size_type) {
case U8_APER_SIZE:
page_order = A_SIZE_8(temp)->page_order;
break;
case U16_APER_SIZE:
page_order = A_SIZE_16(temp)->page_order;
break;
case U32_APER_SIZE:
page_order = A_SIZE_32(temp)->page_order;
break;
case FIXED_APER_SIZE:
page_order = A_SIZE_FIX(temp)->page_order;
break;
case LVL2_APER_SIZE:
/* The generic routines can't deal with 2 level gatt's */
return -EINVAL;
break;
default:
page_order = 0;
break;
}
/* Do not worry about freeing memory, because if this is
* called, then all agp memory is deallocated and removed
* from the table. */
fix historic ioremap() abuse in AGP Several AGP drivers right now use ioremap_nocache() on kernel ram in order to turn a page of regular memory uncached. There are two problems with this: 1) This is a total nightmare for the ioremap() implementation to keep various mappings of the same page coherent. 2) It's a total nightmare for the AGP code since it adds a ton of complexity in terms of keeping track of 2 different pointers to the same thing, in terms of error handling etc etc. This patch fixes this by making the AGP drivers use the new set_memory_XX APIs instead. Note: amd-k7-agp.c is built on Alpha too, and generic.c is built on ia64 as well, which do not yet have the set_memory_*() APIs, so for them some we have a few ugly #ifdefs - hopefully they'll be fixed soon. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Dave Airlie <airlied@linux.ie>
2008-02-06 12:16:00 +08:00
#ifdef CONFIG_X86
set_memory_wb((unsigned long)bridge->gatt_table, 1 << page_order);
#else
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
iounmap(bridge->gatt_table);
fix historic ioremap() abuse in AGP Several AGP drivers right now use ioremap_nocache() on kernel ram in order to turn a page of regular memory uncached. There are two problems with this: 1) This is a total nightmare for the ioremap() implementation to keep various mappings of the same page coherent. 2) It's a total nightmare for the AGP code since it adds a ton of complexity in terms of keeping track of 2 different pointers to the same thing, in terms of error handling etc etc. This patch fixes this by making the AGP drivers use the new set_memory_XX APIs instead. Note: amd-k7-agp.c is built on Alpha too, and generic.c is built on ia64 as well, which do not yet have the set_memory_*() APIs, so for them some we have a few ugly #ifdefs - hopefully they'll be fixed soon. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Dave Airlie <airlied@linux.ie>
2008-02-06 12:16:00 +08:00
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
table = (char *) bridge->gatt_table_real;
table_end = table + ((PAGE_SIZE * (1 << page_order)) - 1);
for (page = virt_to_page(table); page <= virt_to_page(table_end); page++)
ClearPageReserved(page);
[PATCH] AGP fix for Xen VMM When Linux is running on the Xen virtual machine monitor, physical addresses are virtualised and cannot be directly referenced by the AGP GART. This patch fixes the GART driver for Xen by adding a layer of abstraction between physical addresses and 'GART addresses'. Architecture-specific functions are also defined for allocating and freeing the GATT. Xen requires this to ensure that table really is contiguous from the point of view of the GART. These extra interface functions are defined as 'no-ops' for all existing architectures that use the GART driver. Signed-off-by: Keir Fraser <keir@xensource.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Dave Jones <davej@redhat.com>
2005-03-31 05:17:04 +08:00
free_gatt_pages(bridge->gatt_table_real, page_order);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
agp_gatt_table = NULL;
bridge->gatt_table = NULL;
bridge->gatt_table_real = NULL;
bridge->gatt_bus_addr = 0;
return 0;
}
EXPORT_SYMBOL(agp_generic_free_gatt_table);
int agp_generic_insert_memory(struct agp_memory * mem, off_t pg_start, int type)
{
int num_entries;
size_t i;
off_t j;
void *temp;
struct agp_bridge_data *bridge;
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
int mask_type;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
bridge = mem->bridge;
if (!bridge)
return -EINVAL;
[AGPGART] Remove unnecessary flushes when inserting and removing pages. This patch is to speed up flipping of pages in and out of the AGP aperture as needed by the new drm memory manager. A number of global cache flushes are removed as well as some PCI posting flushes. The following guidelines have been used: 1) Memory that is only mapped uncached and that has been subject to a global cache flush after the mapping was changed to uncached does not need any more cache flushes. Neither before binding to the aperture nor after unbinding. 2) Only do one PCI posting flush after a sequence of writes modifying page entries in the GATT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-12-20 23:33:41 +08:00
if (mem->page_count == 0)
return 0;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
temp = bridge->current_size;
switch (bridge->driver->size_type) {
case U8_APER_SIZE:
num_entries = A_SIZE_8(temp)->num_entries;
break;
case U16_APER_SIZE:
num_entries = A_SIZE_16(temp)->num_entries;
break;
case U32_APER_SIZE:
num_entries = A_SIZE_32(temp)->num_entries;
break;
case FIXED_APER_SIZE:
num_entries = A_SIZE_FIX(temp)->num_entries;
break;
case LVL2_APER_SIZE:
/* The generic routines can't deal with 2 level gatt's */
return -EINVAL;
break;
default:
num_entries = 0;
break;
}
num_entries -= agp_memory_reserved/PAGE_SIZE;
if (num_entries < 0) num_entries = 0;
[AGPGART] allow drm populated agp memory types cleanups Fix whitespace, braces, use kzalloc(). Cc: Dave Airlie <airlied@linux.ie> Cc: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Dave Jones <davej@redhat.com>
2007-02-06 08:09:35 +08:00
if (type != mem->type)
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
return -EINVAL;
mask_type = bridge->driver->agp_type_to_mask_type(bridge, type);
if (mask_type != 0) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/* The generic routines know nothing of memory types */
return -EINVAL;
}
/* AK: could wrap */
if ((pg_start + mem->page_count) > num_entries)
return -EINVAL;
j = pg_start;
while (j < (pg_start + mem->page_count)) {
if (!PGE_EMPTY(bridge, readl(bridge->gatt_table+j)))
return -EBUSY;
j++;
}
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
if (!mem->is_flushed) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
bridge->driver->cache_flush();
drivers/char/agp - use bool Use boolean in AGP instead of having own TRUE/FALSE -- Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2008-03-27 05:10:02 +08:00
mem->is_flushed = true;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
for (i = 0, j = pg_start; i < mem->page_count; i++, j++) {
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
writel(bridge->driver->mask_memory(bridge, mem->memory[i], mask_type),
bridge->gatt_table+j);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
[AGPGART] Remove unnecessary flushes when inserting and removing pages. This patch is to speed up flipping of pages in and out of the AGP aperture as needed by the new drm memory manager. A number of global cache flushes are removed as well as some PCI posting flushes. The following guidelines have been used: 1) Memory that is only mapped uncached and that has been subject to a global cache flush after the mapping was changed to uncached does not need any more cache flushes. Neither before binding to the aperture nor after unbinding. 2) Only do one PCI posting flush after a sequence of writes modifying page entries in the GATT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-12-20 23:33:41 +08:00
readl(bridge->gatt_table+j-1); /* PCI Posting. */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
bridge->driver->tlb_flush(mem);
return 0;
}
EXPORT_SYMBOL(agp_generic_insert_memory);
int agp_generic_remove_memory(struct agp_memory *mem, off_t pg_start, int type)
{
size_t i;
struct agp_bridge_data *bridge;
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
int mask_type;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
bridge = mem->bridge;
if (!bridge)
return -EINVAL;
[AGPGART] Remove unnecessary flushes when inserting and removing pages. This patch is to speed up flipping of pages in and out of the AGP aperture as needed by the new drm memory manager. A number of global cache flushes are removed as well as some PCI posting flushes. The following guidelines have been used: 1) Memory that is only mapped uncached and that has been subject to a global cache flush after the mapping was changed to uncached does not need any more cache flushes. Neither before binding to the aperture nor after unbinding. 2) Only do one PCI posting flush after a sequence of writes modifying page entries in the GATT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-12-20 23:33:41 +08:00
if (mem->page_count == 0)
return 0;
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
if (type != mem->type)
return -EINVAL;
mask_type = bridge->driver->agp_type_to_mask_type(bridge, type);
if (mask_type != 0) {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/* The generic routines know nothing of memory types */
return -EINVAL;
}
/* AK: bogus, should encode addresses > 4GB */
for (i = pg_start; i < (mem->page_count + pg_start); i++) {
writel(bridge->scratch_page, bridge->gatt_table+i);
}
[AGPGART] Remove unnecessary flushes when inserting and removing pages. This patch is to speed up flipping of pages in and out of the AGP aperture as needed by the new drm memory manager. A number of global cache flushes are removed as well as some PCI posting flushes. The following guidelines have been used: 1) Memory that is only mapped uncached and that has been subject to a global cache flush after the mapping was changed to uncached does not need any more cache flushes. Neither before binding to the aperture nor after unbinding. 2) Only do one PCI posting flush after a sequence of writes modifying page entries in the GATT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-12-20 23:33:41 +08:00
readl(bridge->gatt_table+i-1); /* PCI Posting. */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
bridge->driver->tlb_flush(mem);
return 0;
}
EXPORT_SYMBOL(agp_generic_remove_memory);
struct agp_memory *agp_generic_alloc_by_type(size_t page_count, int type)
{
return NULL;
}
EXPORT_SYMBOL(agp_generic_alloc_by_type);
void agp_generic_free_by_type(struct agp_memory *curr)
{
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
agp_free_page_array(curr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
agp_free_key(curr->key);
kfree(curr);
}
EXPORT_SYMBOL(agp_generic_free_by_type);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
struct agp_memory *agp_generic_alloc_user(size_t page_count, int type)
{
struct agp_memory *new;
int i;
int pages;
pages = (page_count + ENTRIES_PER_PAGE - 1) / ENTRIES_PER_PAGE;
new = agp_create_user_memory(page_count);
if (new == NULL)
return NULL;
[AGPGART] allow drm populated agp memory types cleanups Fix whitespace, braces, use kzalloc(). Cc: Dave Airlie <airlied@linux.ie> Cc: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Dave Jones <davej@redhat.com>
2007-02-06 08:09:35 +08:00
for (i = 0; i < page_count; i++)
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
new->memory[i] = 0;
new->page_count = 0;
new->type = type;
new->num_scratch_pages = pages;
return new;
}
EXPORT_SYMBOL(agp_generic_alloc_user);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/*
* Basic Page Allocation Routines -
* These routines handle page allocation and by default they reserve the allocated
* memory. They also handle incrementing the current_memory_agp value, Which is checked
* against a maximum value.
*/
void *agp_generic_alloc_page(struct agp_bridge_data *bridge)
{
struct page * page;
[AGP] Allocate AGP pages with GFP_DMA32 by default Not all graphic page remappers support physical addresses over the 4GB mark for remapping, so while some do (the AMD64 GART always did, and I just fixed the i965 to do so properly), we're safest off just forcing GFP_DMA32 allocations to make sure graphics pages get allocated in the low 32-bit address space by default. AGP sub-drivers that really care, and can do better, could just choose to implement their own allocator (or we could add another "64-bit safe" default allocator for their use), but quite frankly, you're not likely to care in practice. So for now, this trivial change means that we won't be allocating pages that we can't map correctly by mistake on x86-64. [ On traditional 32-bit x86, this could never happen, because GFP_KERNEL would never allocate any highmem memory anyway ] Acked-by: Andi Kleen <ak@suse.de> Acked-by: Dave Jones <davej@redhat.com> Cc: Eric Anholt <eric@anholt.net> Cc: Keith Packard <keithp@keithp.com> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-11-23 06:55:29 +08:00
page = alloc_page(GFP_KERNEL | GFP_DMA32);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
if (page == NULL)
return NULL;
map_page_into_agp(page);
get_page(page);
atomic_inc(&agp_bridge->current_memory_agp);
return page_address(page);
}
EXPORT_SYMBOL(agp_generic_alloc_page);
AGP fix race condition between unmapping and freeing pages With Andi's clflush fixup, we were getting hangs on server exit, flushing the mappings after freeing each page helped. This showed up a race condition where the pages after being freed could be reused before the agp mappings had been flushed. Flushing after each single page is a bad thing for future drm work, so make the page destroy a two pass unmapping all the pages, flushing the mappings, and then destroying the pages. Signed-off-by: Dave Airlie <airlied@linux.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-10-15 08:19:16 +08:00
void agp_generic_destroy_page(void *addr, int flags)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
{
struct page *page;
if (addr == NULL)
return;
page = virt_to_page(addr);
AGP fix race condition between unmapping and freeing pages With Andi's clflush fixup, we were getting hangs on server exit, flushing the mappings after freeing each page helped. This showed up a race condition where the pages after being freed could be reused before the agp mappings had been flushed. Flushing after each single page is a bad thing for future drm work, so make the page destroy a two pass unmapping all the pages, flushing the mappings, and then destroying the pages. Signed-off-by: Dave Airlie <airlied@linux.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2007-10-15 08:19:16 +08:00
if (flags & AGP_PAGE_DESTROY_UNMAP)
unmap_page_from_agp(page);
if (flags & AGP_PAGE_DESTROY_FREE) {
put_page(page);
free_page((unsigned long)addr);
atomic_dec(&agp_bridge->current_memory_agp);
}
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
}
EXPORT_SYMBOL(agp_generic_destroy_page);
/* End Basic Page Allocation Routines */
/**
* agp_enable - initialise the agp point-to-point connection.
*
* @mode: agp mode register value to configure with.
*/
void agp_enable(struct agp_bridge_data *bridge, u32 mode)
{
if (!bridge)
return;
bridge->driver->agp_enable(bridge, mode);
}
EXPORT_SYMBOL(agp_enable);
/* When we remove the global variable agp_bridge from all drivers
* then agp_alloc_bridge and agp_generic_find_bridge need to be updated
*/
struct agp_bridge_data *agp_generic_find_bridge(struct pci_dev *pdev)
{
if (list_empty(&agp_bridges))
return NULL;
return agp_bridge;
}
static void ipi_handler(void *null)
{
flush_agp_cache();
}
void global_cache_flush(void)
{
if (on_each_cpu(ipi_handler, NULL, 1, 1) != 0)
panic(PFX "timed out waiting for the other CPUs!\n");
}
EXPORT_SYMBOL(global_cache_flush);
unsigned long agp_generic_mask_memory(struct agp_bridge_data *bridge,
unsigned long addr, int type)
{
/* memory type is ignored in the generic routine */
if (bridge->driver->masks)
return addr | bridge->driver->masks[0].mask;
else
return addr;
}
EXPORT_SYMBOL(agp_generic_mask_memory);
[AGPGART] Allow drm-populated agp memory types This patch allows drm to populate an agpgart structure with pages of its own. It's needed for the new drm memory manager which dynamically flips pages in and out of AGP. The patch modifies the generic functions as well as the intel agp driver. The intel drm driver is currently the only one supporting the new memory manager. Other agp drivers may need some minor fixing up once they have a corresponding memory manager enabled drm driver. AGP memory types >= AGP_USER_TYPES are not populated by the agpgart driver, but the drm is expected to do that, as well as taking care of cache- and tlb flushing when needed. It's not possible to request these types from user space using agpgart ioctls. The Intel driver also gets a new memory type for pages that can be bound cached to the intel GTT. Signed-off-by: Thomas Hellstrom <thomas@tungstengraphics.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-01-23 17:33:43 +08:00
int agp_generic_type_to_mask_type(struct agp_bridge_data *bridge,
int type)
{
if (type >= AGP_USER_TYPES)
return 0;
return type;
}
EXPORT_SYMBOL(agp_generic_type_to_mask_type);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
/*
* These functions are implemented according to the AGPv3 spec,
* which covers implementation details that had previously been
* left open.
*/
int agp3_generic_fetch_size(void)
{
u16 temp_size;
int i;
struct aper_size_info_16 *values;
pci_read_config_word(agp_bridge->dev, agp_bridge->capndx+AGPAPSIZE, &temp_size);
values = A_SIZE_16(agp_bridge->driver->aperture_sizes);
for (i = 0; i < agp_bridge->driver->num_aperture_sizes; i++) {
if (temp_size == values[i].size_value) {
agp_bridge->previous_size =
agp_bridge->current_size = (void *) (values + i);
agp_bridge->aperture_size_idx = i;
return values[i].size;
}
}
return 0;
}
EXPORT_SYMBOL(agp3_generic_fetch_size);
void agp3_generic_tlbflush(struct agp_memory *mem)
{
u32 ctrl;
pci_read_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, &ctrl);
pci_write_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, ctrl & ~AGPCTRL_GTLBEN);
pci_write_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, ctrl);
}
EXPORT_SYMBOL(agp3_generic_tlbflush);
int agp3_generic_configure(void)
{
u32 temp;
struct aper_size_info_16 *current_size;
current_size = A_SIZE_16(agp_bridge->current_size);
pci_read_config_dword(agp_bridge->dev, AGP_APBASE, &temp);
agp_bridge->gart_bus_addr = (temp & PCI_BASE_ADDRESS_MEM_MASK);
/* set aperture size */
pci_write_config_word(agp_bridge->dev, agp_bridge->capndx+AGPAPSIZE, current_size->size_value);
/* set gart pointer */
pci_write_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPGARTLO, agp_bridge->gatt_bus_addr);
/* enable aperture and GTLB */
pci_read_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, &temp);
pci_write_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, temp | AGPCTRL_APERENB | AGPCTRL_GTLBEN);
return 0;
}
EXPORT_SYMBOL(agp3_generic_configure);
void agp3_generic_cleanup(void)
{
u32 ctrl;
pci_read_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, &ctrl);
pci_write_config_dword(agp_bridge->dev, agp_bridge->capndx+AGPCTRL, ctrl & ~AGPCTRL_APERENB);
}
EXPORT_SYMBOL(agp3_generic_cleanup);
[AGPGART] Further constification. Make agp_bridge_driver->aperture_sizes and ->masks const. Also agp_bridge_data->driver Signed-off-by: Dave Jones <davej@redhat.com>
2007-02-23 07:41:28 +08:00
const struct aper_size_info_16 agp3_generic_sizes[AGP_GENERIC_SIZES_ENTRIES] =
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 06:20:36 +08:00
{
{4096, 1048576, 10,0x000},
{2048, 524288, 9, 0x800},
{1024, 262144, 8, 0xc00},
{ 512, 131072, 7, 0xe00},
{ 256, 65536, 6, 0xf00},
{ 128, 32768, 5, 0xf20},
{ 64, 16384, 4, 0xf30},
{ 32, 8192, 3, 0xf38},
{ 16, 4096, 2, 0xf3c},
{ 8, 2048, 1, 0xf3e},
{ 4, 1024, 0, 0xf3f}
};
EXPORT_SYMBOL(agp3_generic_sizes);