2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Re-map IO memory to kernel address space so that we can access it.
|
|
|
|
* This is needed for high PCI addresses that aren't mapped in the
|
|
|
|
* 640k-1MB IO memory area on PC's
|
|
|
|
*
|
|
|
|
* (C) Copyright 1995 1996 Linus Torvalds
|
|
|
|
*/
|
|
|
|
|
2008-01-30 20:34:05 +08:00
|
|
|
#include <linux/bootmem.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <linux/init.h>
|
2006-10-01 14:29:17 +08:00
|
|
|
#include <linux/io.h>
|
2008-01-30 20:34:05 +08:00
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/vmalloc.h>
|
2008-05-13 03:20:57 +08:00
|
|
|
#include <linux/mmiotrace.h>
|
2008-01-30 20:34:05 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <asm/cacheflush.h>
|
2008-01-30 20:34:05 +08:00
|
|
|
#include <asm/e820.h>
|
|
|
|
#include <asm/fixmap.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <asm/pgtable.h>
|
2008-01-30 20:34:05 +08:00
|
|
|
#include <asm/tlbflush.h>
|
2008-01-30 20:34:11 +08:00
|
|
|
#include <asm/pgalloc.h>
|
2008-03-19 08:00:17 +08:00
|
|
|
#include <asm/pat.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-09-11 01:09:38 +08:00
|
|
|
#include "physaddr.h"
|
2008-01-30 20:34:05 +08:00
|
|
|
|
2008-01-30 20:34:05 +08:00
|
|
|
/*
|
|
|
|
* Fix up the linear direct mapping of the kernel to avoid cache attribute
|
|
|
|
* conflicts.
|
|
|
|
*/
|
2008-03-19 08:00:16 +08:00
|
|
|
int ioremap_change_attr(unsigned long vaddr, unsigned long size,
|
2014-11-03 21:01:58 +08:00
|
|
|
enum page_cache_mode pcm)
|
2008-01-30 20:34:05 +08:00
|
|
|
{
|
2008-01-30 20:34:06 +08:00
|
|
|
unsigned long nrpages = size >> PAGE_SHIFT;
|
2008-02-02 00:49:43 +08:00
|
|
|
int err;
|
2008-01-30 20:34:05 +08:00
|
|
|
|
2014-11-03 21:01:58 +08:00
|
|
|
switch (pcm) {
|
|
|
|
case _PAGE_CACHE_MODE_UC:
|
2008-01-30 20:34:06 +08:00
|
|
|
default:
|
2008-03-19 08:00:18 +08:00
|
|
|
err = _set_memory_uc(vaddr, nrpages);
|
2008-01-30 20:34:06 +08:00
|
|
|
break;
|
2014-11-03 21:01:58 +08:00
|
|
|
case _PAGE_CACHE_MODE_WC:
|
2008-03-19 08:00:24 +08:00
|
|
|
err = _set_memory_wc(vaddr, nrpages);
|
|
|
|
break;
|
2015-06-05 00:55:20 +08:00
|
|
|
case _PAGE_CACHE_MODE_WT:
|
|
|
|
err = _set_memory_wt(vaddr, nrpages);
|
|
|
|
break;
|
2014-11-03 21:01:58 +08:00
|
|
|
case _PAGE_CACHE_MODE_WB:
|
2008-03-19 08:00:18 +08:00
|
|
|
err = _set_memory_wb(vaddr, nrpages);
|
2008-01-30 20:34:06 +08:00
|
|
|
break;
|
|
|
|
}
|
2008-01-30 20:34:05 +08:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
x86, ioremap: Speed up check for RAM pages
In __ioremap_caller() (the guts of ioremap), we loop over the range of
pfns being remapped and checks each one individually with page_is_ram().
For large ioremaps, this can be very slow. For example, we have a
device with a 256 GiB PCI BAR, and ioremapping this BAR can take 20+
seconds -- sometimes long enough to trigger the soft lockup detector!
Internally, page_is_ram() calls walk_system_ram_range() on a single
page. Instead, we can make a single call to walk_system_ram_range()
from __ioremap_caller(), and do our further checks only for any RAM
pages that we find. For the common case of MMIO, this saves an enormous
amount of work, since the range being ioremapped doesn't intersect
system RAM at all.
With this change, ioremap on our 256 GiB BAR takes less than 1 second.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Link: http://lkml.kernel.org/r/1399054721-1331-1-git-send-email-roland@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2014-05-03 02:18:41 +08:00
|
|
|
static int __ioremap_check_ram(unsigned long start_pfn, unsigned long nr_pages,
|
|
|
|
void *arg)
|
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
|
|
|
|
for (i = 0; i < nr_pages; ++i)
|
|
|
|
if (pfn_valid(start_pfn + i) &&
|
|
|
|
!PageReserved(pfn_to_page(start_pfn + i)))
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Remap an arbitrary physical address space into the kernel virtual
|
2015-04-15 06:47:29 +08:00
|
|
|
* address space. It transparently creates kernel huge I/O mapping when
|
|
|
|
* the physical address is aligned by a huge page size (1GB or 2MB) and
|
|
|
|
* the requested size is at least the huge page size.
|
|
|
|
*
|
|
|
|
* NOTE: MTRRs can override PAT memory types with a 4KB granularity.
|
|
|
|
* Therefore, the mapping code falls back to use a smaller page toward 4KB
|
|
|
|
* when a mapping range is covered by non-WB type of MTRRs.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* NOTE! We need to allow non-page-aligned mappings too: we will obviously
|
|
|
|
* have to convert them into an offset in a page-aligned mapping, but the
|
|
|
|
* caller shouldn't need to know that small detail.
|
|
|
|
*/
|
2008-04-28 17:12:42 +08:00
|
|
|
static void __iomem *__ioremap_caller(resource_size_t phys_addr,
|
2014-11-03 21:01:58 +08:00
|
|
|
unsigned long size, enum page_cache_mode pcm, void *caller)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2010-06-18 11:22:40 +08:00
|
|
|
unsigned long offset, vaddr;
|
|
|
|
resource_size_t pfn, last_pfn, last_addr;
|
2008-05-13 03:21:03 +08:00
|
|
|
const resource_size_t unaligned_phys_addr = phys_addr;
|
|
|
|
const unsigned long unaligned_size = size;
|
2008-01-30 20:34:05 +08:00
|
|
|
struct vm_struct *area;
|
2014-11-03 21:01:58 +08:00
|
|
|
enum page_cache_mode new_pcm;
|
2008-01-30 20:34:06 +08:00
|
|
|
pgprot_t prot;
|
2008-03-25 05:39:55 +08:00
|
|
|
int retval;
|
2008-05-13 03:20:57 +08:00
|
|
|
void __iomem *ret_addr;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Don't allow wraparound or zero size */
|
|
|
|
last_addr = phys_addr + size - 1;
|
|
|
|
if (!size || last_addr < phys_addr)
|
|
|
|
return NULL;
|
|
|
|
|
2008-02-28 03:57:40 +08:00
|
|
|
if (!phys_addr_valid(phys_addr)) {
|
2008-03-19 08:00:25 +08:00
|
|
|
printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
|
2008-04-11 06:09:50 +08:00
|
|
|
(unsigned long long)phys_addr);
|
2008-02-28 03:57:40 +08:00
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Don't remap the low PCI/ISA area, it's always mapped..
|
|
|
|
*/
|
2008-06-21 03:58:46 +08:00
|
|
|
if (is_ISA_range(phys_addr, last_addr))
|
2008-01-30 20:34:05 +08:00
|
|
|
return (__force void __iomem *)phys_to_virt(phys_addr);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't allow anybody to remap normal RAM that we're using..
|
|
|
|
*/
|
2015-07-17 07:23:15 +08:00
|
|
|
pfn = phys_addr >> PAGE_SHIFT;
|
|
|
|
last_pfn = last_addr >> PAGE_SHIFT;
|
|
|
|
if (walk_system_ram_range(pfn, last_pfn - pfn + 1, NULL,
|
2015-07-17 07:23:14 +08:00
|
|
|
__ioremap_check_ram) == 1) {
|
2015-07-24 22:13:43 +08:00
|
|
|
WARN_ONCE(1, "ioremap on RAM at %pa - %pa\n",
|
|
|
|
&phys_addr, &last_addr);
|
2015-07-17 07:23:15 +08:00
|
|
|
return NULL;
|
2014-10-14 06:54:05 +08:00
|
|
|
}
|
2015-07-17 07:23:15 +08:00
|
|
|
|
2008-03-19 08:00:17 +08:00
|
|
|
/*
|
|
|
|
* Mappings have to be page-aligned
|
|
|
|
*/
|
|
|
|
offset = phys_addr & ~PAGE_MASK;
|
2010-06-18 11:22:40 +08:00
|
|
|
phys_addr &= PHYSICAL_PAGE_MASK;
|
2008-03-19 08:00:17 +08:00
|
|
|
size = PAGE_ALIGN(last_addr+1) - phys_addr;
|
|
|
|
|
2008-08-16 00:12:47 +08:00
|
|
|
retval = reserve_memtype(phys_addr, (u64)phys_addr + size,
|
2014-11-03 21:01:59 +08:00
|
|
|
pcm, &new_pcm);
|
2008-03-25 05:39:55 +08:00
|
|
|
if (retval) {
|
2009-07-11 00:57:33 +08:00
|
|
|
printk(KERN_ERR "ioremap reserve_memtype failed %d\n", retval);
|
2008-03-25 05:39:55 +08:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-11-03 21:01:58 +08:00
|
|
|
if (pcm != new_pcm) {
|
|
|
|
if (!is_new_memtype_allowed(phys_addr, size, pcm, new_pcm)) {
|
2009-07-11 00:57:33 +08:00
|
|
|
printk(KERN_ERR
|
2014-11-03 21:01:58 +08:00
|
|
|
"ioremap error for 0x%llx-0x%llx, requested 0x%x, got 0x%x\n",
|
2008-04-11 06:09:50 +08:00
|
|
|
(unsigned long long)phys_addr,
|
|
|
|
(unsigned long long)(phys_addr + size),
|
2014-11-03 21:01:58 +08:00
|
|
|
pcm, new_pcm);
|
2009-11-05 10:43:51 +08:00
|
|
|
goto err_free_memtype;
|
2008-03-19 08:00:17 +08:00
|
|
|
}
|
2014-11-03 21:01:58 +08:00
|
|
|
pcm = new_pcm;
|
2008-03-19 08:00:17 +08:00
|
|
|
}
|
|
|
|
|
2014-11-03 21:01:58 +08:00
|
|
|
prot = PAGE_KERNEL_IO;
|
|
|
|
switch (pcm) {
|
|
|
|
case _PAGE_CACHE_MODE_UC:
|
2008-01-30 20:34:06 +08:00
|
|
|
default:
|
2014-11-03 21:01:58 +08:00
|
|
|
prot = __pgprot(pgprot_val(prot) |
|
|
|
|
cachemode2protval(_PAGE_CACHE_MODE_UC));
|
2008-01-30 20:34:06 +08:00
|
|
|
break;
|
2014-11-03 21:01:58 +08:00
|
|
|
case _PAGE_CACHE_MODE_UC_MINUS:
|
|
|
|
prot = __pgprot(pgprot_val(prot) |
|
|
|
|
cachemode2protval(_PAGE_CACHE_MODE_UC_MINUS));
|
2008-04-26 08:07:22 +08:00
|
|
|
break;
|
2014-11-03 21:01:58 +08:00
|
|
|
case _PAGE_CACHE_MODE_WC:
|
|
|
|
prot = __pgprot(pgprot_val(prot) |
|
|
|
|
cachemode2protval(_PAGE_CACHE_MODE_WC));
|
2008-03-19 08:00:24 +08:00
|
|
|
break;
|
2015-06-05 00:55:15 +08:00
|
|
|
case _PAGE_CACHE_MODE_WT:
|
|
|
|
prot = __pgprot(pgprot_val(prot) |
|
|
|
|
cachemode2protval(_PAGE_CACHE_MODE_WT));
|
|
|
|
break;
|
2014-11-03 21:01:58 +08:00
|
|
|
case _PAGE_CACHE_MODE_WB:
|
2008-01-30 20:34:06 +08:00
|
|
|
break;
|
|
|
|
}
|
2006-10-01 14:29:17 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Ok, go for it..
|
|
|
|
*/
|
2008-04-28 17:12:42 +08:00
|
|
|
area = get_vm_area_caller(size, VM_IOREMAP, caller);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (!area)
|
2009-11-05 10:43:51 +08:00
|
|
|
goto err_free_memtype;
|
2005-04-17 06:20:36 +08:00
|
|
|
area->phys_addr = phys_addr;
|
2008-02-04 23:48:05 +08:00
|
|
|
vaddr = (unsigned long) area->addr;
|
2009-04-10 05:26:47 +08:00
|
|
|
|
2014-11-03 21:01:58 +08:00
|
|
|
if (kernel_map_sync_memtype(phys_addr, size, pcm))
|
2009-11-05 10:43:51 +08:00
|
|
|
goto err_free_area;
|
2008-01-30 20:34:05 +08:00
|
|
|
|
2009-11-05 10:43:51 +08:00
|
|
|
if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot))
|
|
|
|
goto err_free_area;
|
2008-01-30 20:34:05 +08:00
|
|
|
|
2008-05-13 03:20:57 +08:00
|
|
|
ret_addr = (void __iomem *) (vaddr + offset);
|
2008-05-13 03:21:03 +08:00
|
|
|
mmiotrace_ioremap(unaligned_phys_addr, unaligned_size, ret_addr);
|
2008-05-13 03:20:57 +08:00
|
|
|
|
2011-04-29 01:00:30 +08:00
|
|
|
/*
|
|
|
|
* Check if the request spans more than any BAR in the iomem resource
|
|
|
|
* tree.
|
|
|
|
*/
|
2015-12-22 04:01:14 +08:00
|
|
|
if (iomem_map_sanity_check(unaligned_phys_addr, unaligned_size))
|
|
|
|
pr_warn("caller %pS mapping multiple BARs\n", caller);
|
2011-04-29 01:00:30 +08:00
|
|
|
|
2008-05-13 03:20:57 +08:00
|
|
|
return ret_addr;
|
2009-11-05 10:43:51 +08:00
|
|
|
err_free_area:
|
|
|
|
free_vm_area(area);
|
|
|
|
err_free_memtype:
|
|
|
|
free_memtype(phys_addr, phys_addr + size);
|
|
|
|
return NULL;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ioremap_nocache - map bus memory into CPU space
|
2012-06-10 10:50:52 +08:00
|
|
|
* @phys_addr: bus address of the memory
|
2005-04-17 06:20:36 +08:00
|
|
|
* @size: size of the resource to map
|
|
|
|
*
|
|
|
|
* ioremap_nocache performs a platform specific sequence of operations to
|
|
|
|
* make bus memory CPU accessible via the readb/readw/readl/writeb/
|
|
|
|
* writew/writel functions and the other mmio helpers. The returned
|
|
|
|
* address is not guaranteed to be usable directly as a virtual
|
2008-01-30 20:34:05 +08:00
|
|
|
* address.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* This version of ioremap ensures that the memory is marked uncachable
|
|
|
|
* on the CPU as well as honouring existing caching rules from things like
|
2008-01-30 20:34:05 +08:00
|
|
|
* the PCI bus. Note that there are other caches and buffers on many
|
2005-04-17 06:20:36 +08:00
|
|
|
* busses. In particular driver authors should read up on PCI writes
|
|
|
|
*
|
|
|
|
* It's useful if some control registers are in such an area and
|
|
|
|
* write combining or read caching is not desirable:
|
2008-01-30 20:34:05 +08:00
|
|
|
*
|
2005-04-17 06:20:36 +08:00
|
|
|
* Must be freed with iounmap.
|
|
|
|
*/
|
2008-03-25 02:22:39 +08:00
|
|
|
void __iomem *ioremap_nocache(resource_size_t phys_addr, unsigned long size)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2008-04-26 08:07:22 +08:00
|
|
|
/*
|
|
|
|
* Ideally, this should be:
|
2015-05-26 16:28:15 +08:00
|
|
|
* pat_enabled() ? _PAGE_CACHE_MODE_UC : _PAGE_CACHE_MODE_UC_MINUS;
|
2008-04-26 08:07:22 +08:00
|
|
|
*
|
|
|
|
* Till we fix all X drivers to use ioremap_wc(), we will use
|
x86/mm: Add ioremap_uc() helper to map memory uncacheable (not UC-)
ioremap_nocache() currently uses UC- by default. Our goal is to
eventually make UC the default. Linux maps UC- to PCD=1, PWT=0
page attributes on non-PAT systems. Linux maps UC to PCD=1,
PWT=1 page attributes on non-PAT systems. On non-PAT and PAT
systems a WC MTRR has different effects on pages with either of
these attributes. In order to help with a smooth transition its
best to enable use of UC (PCD,1, PWT=1) on a region as that
ensures a WC MTRR will have no effect on a region, this however
requires us to have an way to declare a region as UC and we
currently do not have a way to do this.
WC MTRR on non-PAT system with PCD=1, PWT=0 (UC-) yields WC.
WC MTRR on non-PAT system with PCD=1, PWT=1 (UC) yields UC.
WC MTRR on PAT system with PCD=1, PWT=0 (UC-) yields WC.
WC MTRR on PAT system with PCD=1, PWT=1 (UC) yields UC.
A flip of the default ioremap_nocache() behaviour from UC- to UC
can therefore regress a memory region from effective memory type
WC to UC if MTRRs are used. Use of MTRRs should be phased out
and in the best case only arch_phys_wc_add() use will remain,
even if this happens arch_phys_wc_add() will have an effect on
non-PAT systems and changes to default ioremap_nocache()
behaviour could regress drivers.
Now, ideally we'd use ioremap_nocache() on the regions in which
we'd need uncachable memory types and avoid any MTRRs on those
regions. There are however some restrictions on MTRRs use, such
as the requirement of having the base and size of variable sized
MTRRs to be powers of two, which could mean having to use a WC
MTRR over a large area which includes a region in which
write-combining effects are undesirable.
Add ioremap_uc() to help with the both phasing out of MTRR use
and also provide a way to blacklist small WC undesirable regions
in devices with mixed regions which are size-implicated to use
large WC MTRRs. Use of ioremap_uc() helps phase out MTRR use by
avoiding regressions with an eventual flip of default behaviour
or ioremap_nocache() from UC- to UC.
Drivers working with WC MTRRs can use the below table to review
and consider the use of ioremap*() and similar helpers to ensure
appropriate behaviour long term even if default
ioremap_nocache() behaviour changes from UC- to UC.
Although ioremap_uc() is being added we leave set_memory_uc() to
use UC- as only initial memory type setup is required to be able
to accommodate existing device drivers and phase out MTRR use.
It should also be clarified that set_memory_uc() cannot be used
with IO memory, even though its use will not return any errors,
it really has no effect.
----------------------------------------------------------------------
MTRR Non-PAT PAT Linux ioremap value Effective memory type
----------------------------------------------------------------------
Non-PAT | PAT
PAT
|PCD
||PWT
|||
WC 000 WB _PAGE_CACHE_MODE_WB WC | WC
WC 001 WC _PAGE_CACHE_MODE_WC WC* | WC
WC 010 UC- _PAGE_CACHE_MODE_UC_MINUS WC* | WC
WC 011 UC _PAGE_CACHE_MODE_UC UC | UC
----------------------------------------------------------------------
Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Antonino Daplas <adaplas@gmail.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Mike Travis <travis@sgi.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suresh Siddha <sbsiddha@gmail.com>
Cc: Thierry Reding <treding@nvidia.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Ville Syrjälä <syrjala@sci.fi>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-fbdev@vger.kernel.org
Link: http://lkml.kernel.org/r/1430343851-967-2-git-send-email-mcgrof@do-not-panic.com
Link: http://lkml.kernel.org/r/1431332153-18566-9-git-send-email-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 16:15:53 +08:00
|
|
|
* UC MINUS. Drivers that are certain they need or can already
|
|
|
|
* be converted over to strong UC can use ioremap_uc().
|
2008-04-26 08:07:22 +08:00
|
|
|
*/
|
2014-11-03 21:01:58 +08:00
|
|
|
enum page_cache_mode pcm = _PAGE_CACHE_MODE_UC_MINUS;
|
2008-04-26 08:07:22 +08:00
|
|
|
|
2014-11-03 21:01:58 +08:00
|
|
|
return __ioremap_caller(phys_addr, size, pcm,
|
2008-04-28 17:12:42 +08:00
|
|
|
__builtin_return_address(0));
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2005-06-23 15:08:33 +08:00
|
|
|
EXPORT_SYMBOL(ioremap_nocache);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
x86/mm: Add ioremap_uc() helper to map memory uncacheable (not UC-)
ioremap_nocache() currently uses UC- by default. Our goal is to
eventually make UC the default. Linux maps UC- to PCD=1, PWT=0
page attributes on non-PAT systems. Linux maps UC to PCD=1,
PWT=1 page attributes on non-PAT systems. On non-PAT and PAT
systems a WC MTRR has different effects on pages with either of
these attributes. In order to help with a smooth transition its
best to enable use of UC (PCD,1, PWT=1) on a region as that
ensures a WC MTRR will have no effect on a region, this however
requires us to have an way to declare a region as UC and we
currently do not have a way to do this.
WC MTRR on non-PAT system with PCD=1, PWT=0 (UC-) yields WC.
WC MTRR on non-PAT system with PCD=1, PWT=1 (UC) yields UC.
WC MTRR on PAT system with PCD=1, PWT=0 (UC-) yields WC.
WC MTRR on PAT system with PCD=1, PWT=1 (UC) yields UC.
A flip of the default ioremap_nocache() behaviour from UC- to UC
can therefore regress a memory region from effective memory type
WC to UC if MTRRs are used. Use of MTRRs should be phased out
and in the best case only arch_phys_wc_add() use will remain,
even if this happens arch_phys_wc_add() will have an effect on
non-PAT systems and changes to default ioremap_nocache()
behaviour could regress drivers.
Now, ideally we'd use ioremap_nocache() on the regions in which
we'd need uncachable memory types and avoid any MTRRs on those
regions. There are however some restrictions on MTRRs use, such
as the requirement of having the base and size of variable sized
MTRRs to be powers of two, which could mean having to use a WC
MTRR over a large area which includes a region in which
write-combining effects are undesirable.
Add ioremap_uc() to help with the both phasing out of MTRR use
and also provide a way to blacklist small WC undesirable regions
in devices with mixed regions which are size-implicated to use
large WC MTRRs. Use of ioremap_uc() helps phase out MTRR use by
avoiding regressions with an eventual flip of default behaviour
or ioremap_nocache() from UC- to UC.
Drivers working with WC MTRRs can use the below table to review
and consider the use of ioremap*() and similar helpers to ensure
appropriate behaviour long term even if default
ioremap_nocache() behaviour changes from UC- to UC.
Although ioremap_uc() is being added we leave set_memory_uc() to
use UC- as only initial memory type setup is required to be able
to accommodate existing device drivers and phase out MTRR use.
It should also be clarified that set_memory_uc() cannot be used
with IO memory, even though its use will not return any errors,
it really has no effect.
----------------------------------------------------------------------
MTRR Non-PAT PAT Linux ioremap value Effective memory type
----------------------------------------------------------------------
Non-PAT | PAT
PAT
|PCD
||PWT
|||
WC 000 WB _PAGE_CACHE_MODE_WB WC | WC
WC 001 WC _PAGE_CACHE_MODE_WC WC* | WC
WC 010 UC- _PAGE_CACHE_MODE_UC_MINUS WC* | WC
WC 011 UC _PAGE_CACHE_MODE_UC UC | UC
----------------------------------------------------------------------
Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Antonino Daplas <adaplas@gmail.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Mike Travis <travis@sgi.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suresh Siddha <sbsiddha@gmail.com>
Cc: Thierry Reding <treding@nvidia.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Ville Syrjälä <syrjala@sci.fi>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-fbdev@vger.kernel.org
Link: http://lkml.kernel.org/r/1430343851-967-2-git-send-email-mcgrof@do-not-panic.com
Link: http://lkml.kernel.org/r/1431332153-18566-9-git-send-email-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2015-05-11 16:15:53 +08:00
|
|
|
/**
|
|
|
|
* ioremap_uc - map bus memory into CPU space as strongly uncachable
|
|
|
|
* @phys_addr: bus address of the memory
|
|
|
|
* @size: size of the resource to map
|
|
|
|
*
|
|
|
|
* ioremap_uc performs a platform specific sequence of operations to
|
|
|
|
* make bus memory CPU accessible via the readb/readw/readl/writeb/
|
|
|
|
* writew/writel functions and the other mmio helpers. The returned
|
|
|
|
* address is not guaranteed to be usable directly as a virtual
|
|
|
|
* address.
|
|
|
|
*
|
|
|
|
* This version of ioremap ensures that the memory is marked with a strong
|
|
|
|
* preference as completely uncachable on the CPU when possible. For non-PAT
|
|
|
|
* systems this ends up setting page-attribute flags PCD=1, PWT=1. For PAT
|
|
|
|
* systems this will set the PAT entry for the pages as strong UC. This call
|
|
|
|
* will honor existing caching rules from things like the PCI bus. Note that
|
|
|
|
* there are other caches and buffers on many busses. In particular driver
|
|
|
|
* authors should read up on PCI writes.
|
|
|
|
*
|
|
|
|
* It's useful if some control registers are in such an area and
|
|
|
|
* write combining or read caching is not desirable:
|
|
|
|
*
|
|
|
|
* Must be freed with iounmap.
|
|
|
|
*/
|
|
|
|
void __iomem *ioremap_uc(resource_size_t phys_addr, unsigned long size)
|
|
|
|
{
|
|
|
|
enum page_cache_mode pcm = _PAGE_CACHE_MODE_UC;
|
|
|
|
|
|
|
|
return __ioremap_caller(phys_addr, size, pcm,
|
|
|
|
__builtin_return_address(0));
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(ioremap_uc);
|
|
|
|
|
2008-03-19 08:00:24 +08:00
|
|
|
/**
|
|
|
|
* ioremap_wc - map memory into CPU space write combined
|
2012-06-10 10:50:52 +08:00
|
|
|
* @phys_addr: bus address of the memory
|
2008-03-19 08:00:24 +08:00
|
|
|
* @size: size of the resource to map
|
|
|
|
*
|
|
|
|
* This version of ioremap ensures that the memory is marked write combining.
|
|
|
|
* Write combining allows faster writes to some hardware devices.
|
|
|
|
*
|
|
|
|
* Must be freed with iounmap.
|
|
|
|
*/
|
2009-01-10 08:13:13 +08:00
|
|
|
void __iomem *ioremap_wc(resource_size_t phys_addr, unsigned long size)
|
2008-03-19 08:00:24 +08:00
|
|
|
{
|
2015-06-05 00:55:11 +08:00
|
|
|
return __ioremap_caller(phys_addr, size, _PAGE_CACHE_MODE_WC,
|
2008-04-28 17:12:42 +08:00
|
|
|
__builtin_return_address(0));
|
2008-03-19 08:00:24 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(ioremap_wc);
|
|
|
|
|
2015-06-05 00:55:15 +08:00
|
|
|
/**
|
|
|
|
* ioremap_wt - map memory into CPU space write through
|
|
|
|
* @phys_addr: bus address of the memory
|
|
|
|
* @size: size of the resource to map
|
|
|
|
*
|
|
|
|
* This version of ioremap ensures that the memory is marked write through.
|
|
|
|
* Write through stores data into memory while keeping the cache up-to-date.
|
|
|
|
*
|
|
|
|
* Must be freed with iounmap.
|
|
|
|
*/
|
|
|
|
void __iomem *ioremap_wt(resource_size_t phys_addr, unsigned long size)
|
|
|
|
{
|
|
|
|
return __ioremap_caller(phys_addr, size, _PAGE_CACHE_MODE_WT,
|
|
|
|
__builtin_return_address(0));
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(ioremap_wt);
|
|
|
|
|
2008-03-25 02:22:39 +08:00
|
|
|
void __iomem *ioremap_cache(resource_size_t phys_addr, unsigned long size)
|
2008-01-30 20:34:06 +08:00
|
|
|
{
|
2014-11-03 21:01:58 +08:00
|
|
|
return __ioremap_caller(phys_addr, size, _PAGE_CACHE_MODE_WB,
|
2008-04-28 17:12:42 +08:00
|
|
|
__builtin_return_address(0));
|
2008-01-30 20:34:06 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(ioremap_cache);
|
|
|
|
|
2008-07-24 12:27:05 +08:00
|
|
|
void __iomem *ioremap_prot(resource_size_t phys_addr, unsigned long size,
|
|
|
|
unsigned long prot_val)
|
|
|
|
{
|
2014-11-03 21:01:58 +08:00
|
|
|
return __ioremap_caller(phys_addr, size,
|
|
|
|
pgprot2cachemode(__pgprot(prot_val)),
|
2008-07-24 12:27:05 +08:00
|
|
|
__builtin_return_address(0));
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(ioremap_prot);
|
|
|
|
|
2005-12-13 14:17:09 +08:00
|
|
|
/**
|
|
|
|
* iounmap - Free a IO remapping
|
|
|
|
* @addr: virtual address from ioremap_*
|
|
|
|
*
|
|
|
|
* Caller must ensure there is only one unmapping for the same pointer.
|
|
|
|
*/
|
2005-04-17 06:20:36 +08:00
|
|
|
void iounmap(volatile void __iomem *addr)
|
|
|
|
{
|
2005-12-13 14:17:09 +08:00
|
|
|
struct vm_struct *p, *o;
|
2005-07-08 08:56:02 +08:00
|
|
|
|
|
|
|
if ((void __force *)addr <= high_memory)
|
2005-04-17 06:20:36 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* __ioremap special-cases the PCI/ISA range by not instantiating a
|
|
|
|
* vm_area and by simply returning an address into the kernel mapping
|
|
|
|
* of ISA space. So handle that here.
|
|
|
|
*/
|
2008-05-12 21:43:35 +08:00
|
|
|
if ((void __force *)addr >= phys_to_virt(ISA_START_ADDRESS) &&
|
|
|
|
(void __force *)addr < phys_to_virt(ISA_END_ADDRESS))
|
2005-04-17 06:20:36 +08:00
|
|
|
return;
|
|
|
|
|
2008-01-30 20:34:05 +08:00
|
|
|
addr = (volatile void __iomem *)
|
|
|
|
(PAGE_MASK & (unsigned long __force)addr);
|
2005-12-13 14:17:09 +08:00
|
|
|
|
2008-05-13 03:20:57 +08:00
|
|
|
mmiotrace_iounmap(addr);
|
|
|
|
|
2005-12-13 14:17:09 +08:00
|
|
|
/* Use the vm area unlocked, assuming the caller
|
|
|
|
ensures there isn't another iounmap for the same address
|
|
|
|
in parallel. Reuse of the virtual address is prevented by
|
|
|
|
leaving it in the global lists until we're done with it.
|
|
|
|
cpa takes care of the direct mappings. */
|
2013-04-30 06:07:27 +08:00
|
|
|
p = find_vm_area((void __force *)addr);
|
2005-12-13 14:17:09 +08:00
|
|
|
|
|
|
|
if (!p) {
|
2008-01-30 20:34:05 +08:00
|
|
|
printk(KERN_ERR "iounmap: bad address %p\n", addr);
|
2005-07-08 08:56:02 +08:00
|
|
|
dump_stack();
|
2005-12-13 14:17:09 +08:00
|
|
|
return;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2008-03-19 08:00:17 +08:00
|
|
|
free_memtype(p->phys_addr, p->phys_addr + get_vm_area_size(p));
|
|
|
|
|
2005-12-13 14:17:09 +08:00
|
|
|
/* Finally remove it */
|
2008-05-12 21:43:35 +08:00
|
|
|
o = remove_vm_area((void __force *)addr);
|
2005-12-13 14:17:09 +08:00
|
|
|
BUG_ON(p != o || o == NULL);
|
2008-01-30 20:34:05 +08:00
|
|
|
kfree(p);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2005-06-23 15:08:33 +08:00
|
|
|
EXPORT_SYMBOL(iounmap);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2015-05-28 16:29:27 +08:00
|
|
|
int __init arch_ioremap_pud_supported(void)
|
2015-04-15 06:47:29 +08:00
|
|
|
{
|
|
|
|
#ifdef CONFIG_X86_64
|
2016-03-29 23:41:58 +08:00
|
|
|
return boot_cpu_has(X86_FEATURE_GBPAGES);
|
2015-04-15 06:47:29 +08:00
|
|
|
#else
|
|
|
|
return 0;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2015-05-28 16:29:27 +08:00
|
|
|
int __init arch_ioremap_pmd_supported(void)
|
2015-04-15 06:47:29 +08:00
|
|
|
{
|
2016-03-29 23:42:03 +08:00
|
|
|
return boot_cpu_has(X86_FEATURE_PSE);
|
2015-04-15 06:47:29 +08:00
|
|
|
}
|
|
|
|
|
2008-03-19 08:00:15 +08:00
|
|
|
/*
|
|
|
|
* Convert a physical pointer to a virtual kernel pointer for /dev/mem
|
|
|
|
* access
|
|
|
|
*/
|
2014-07-28 23:20:33 +08:00
|
|
|
void *xlate_dev_mem_ptr(phys_addr_t phys)
|
2008-03-19 08:00:15 +08:00
|
|
|
{
|
2012-11-24 02:19:07 +08:00
|
|
|
unsigned long start = phys & PAGE_MASK;
|
|
|
|
unsigned long offset = phys & ~PAGE_MASK;
|
2015-05-08 18:43:53 +08:00
|
|
|
void *vaddr;
|
2008-03-19 08:00:15 +08:00
|
|
|
|
|
|
|
/* If page is RAM, we can use __va. Otherwise ioremap and unmap. */
|
|
|
|
if (page_is_ram(start >> PAGE_SHIFT))
|
|
|
|
return __va(phys);
|
|
|
|
|
2015-05-08 18:43:53 +08:00
|
|
|
vaddr = ioremap_cache(start, PAGE_SIZE);
|
2012-11-24 02:19:07 +08:00
|
|
|
/* Only add the offset on success and return NULL if the ioremap() failed: */
|
|
|
|
if (vaddr)
|
|
|
|
vaddr += offset;
|
2008-03-19 08:00:15 +08:00
|
|
|
|
2015-05-08 18:43:53 +08:00
|
|
|
return vaddr;
|
2008-03-19 08:00:15 +08:00
|
|
|
}
|
|
|
|
|
2014-07-28 23:20:33 +08:00
|
|
|
void unxlate_dev_mem_ptr(phys_addr_t phys, void *addr)
|
2008-03-19 08:00:15 +08:00
|
|
|
{
|
|
|
|
if (page_is_ram(phys >> PAGE_SHIFT))
|
|
|
|
return;
|
|
|
|
|
|
|
|
iounmap((void __iomem *)((unsigned long)addr & PAGE_MASK));
|
|
|
|
}
|
|
|
|
|
2009-03-21 08:53:34 +08:00
|
|
|
static pte_t bm_pte[PAGE_SIZE/sizeof(pte_t)] __page_aligned_bss;
|
2008-01-30 20:33:44 +08:00
|
|
|
|
2008-02-10 06:24:09 +08:00
|
|
|
static inline pmd_t * __init early_ioremap_pmd(unsigned long addr)
|
2008-01-30 20:33:44 +08:00
|
|
|
{
|
2008-02-13 23:20:35 +08:00
|
|
|
/* Don't assume we're using swapper_pg_dir at this point */
|
|
|
|
pgd_t *base = __va(read_cr3());
|
|
|
|
pgd_t *pgd = &base[pgd_index(addr)];
|
2008-02-10 06:24:09 +08:00
|
|
|
pud_t *pud = pud_offset(pgd, addr);
|
|
|
|
pmd_t *pmd = pmd_offset(pud, addr);
|
|
|
|
|
|
|
|
return pmd;
|
2008-01-30 20:33:44 +08:00
|
|
|
}
|
|
|
|
|
2008-02-10 06:24:09 +08:00
|
|
|
static inline pte_t * __init early_ioremap_pte(unsigned long addr)
|
2008-01-30 20:33:44 +08:00
|
|
|
{
|
2008-02-10 06:24:09 +08:00
|
|
|
return &bm_pte[pte_index(addr)];
|
2008-01-30 20:33:44 +08:00
|
|
|
}
|
|
|
|
|
2010-10-14 07:02:24 +08:00
|
|
|
bool __init is_early_ioremap_ptep(pte_t *ptep)
|
|
|
|
{
|
|
|
|
return ptep >= &bm_pte[0] && ptep < &bm_pte[PAGE_SIZE/sizeof(pte_t)];
|
|
|
|
}
|
|
|
|
|
2008-01-30 20:33:44 +08:00
|
|
|
void __init early_ioremap_init(void)
|
2008-01-30 20:33:44 +08:00
|
|
|
{
|
2008-02-10 06:24:09 +08:00
|
|
|
pmd_t *pmd;
|
2008-01-30 20:33:44 +08:00
|
|
|
|
2014-05-06 03:19:31 +08:00
|
|
|
#ifdef CONFIG_X86_64
|
|
|
|
BUILD_BUG_ON((fix_to_virt(0) + PAGE_SIZE) & ((1 << PMD_SHIFT) - 1));
|
|
|
|
#else
|
|
|
|
WARN_ON((fix_to_virt(0) + PAGE_SIZE) & ((1 << PMD_SHIFT) - 1));
|
|
|
|
#endif
|
|
|
|
|
2014-04-08 06:39:49 +08:00
|
|
|
early_ioremap_setup();
|
2009-03-07 13:34:19 +08:00
|
|
|
|
2008-02-10 06:24:09 +08:00
|
|
|
pmd = early_ioremap_pmd(fix_to_virt(FIX_BTMAP_BEGIN));
|
2009-03-21 08:53:34 +08:00
|
|
|
memset(bm_pte, 0, sizeof(bm_pte));
|
|
|
|
pmd_populate_kernel(&init_mm, pmd, bm_pte);
|
2008-02-10 06:24:09 +08:00
|
|
|
|
2008-01-30 20:33:49 +08:00
|
|
|
/*
|
2008-02-10 06:24:09 +08:00
|
|
|
* The boot-ioremap range spans multiple pmds, for which
|
2008-01-30 20:33:49 +08:00
|
|
|
* we are not prepared:
|
|
|
|
*/
|
2009-12-19 00:05:51 +08:00
|
|
|
#define __FIXADDR_TOP (-PAGE_SIZE)
|
|
|
|
BUILD_BUG_ON((__fix_to_virt(FIX_BTMAP_BEGIN) >> PMD_SHIFT)
|
|
|
|
!= (__fix_to_virt(FIX_BTMAP_END) >> PMD_SHIFT));
|
|
|
|
#undef __FIXADDR_TOP
|
2008-02-10 06:24:09 +08:00
|
|
|
if (pmd != early_ioremap_pmd(fix_to_virt(FIX_BTMAP_END))) {
|
2008-01-30 20:33:49 +08:00
|
|
|
WARN_ON(1);
|
2008-02-10 06:24:09 +08:00
|
|
|
printk(KERN_WARNING "pmd %p != %p\n",
|
|
|
|
pmd, early_ioremap_pmd(fix_to_virt(FIX_BTMAP_END)));
|
2008-01-30 20:34:05 +08:00
|
|
|
printk(KERN_WARNING "fix_to_virt(FIX_BTMAP_BEGIN): %08lx\n",
|
2008-02-10 06:24:09 +08:00
|
|
|
fix_to_virt(FIX_BTMAP_BEGIN));
|
2008-01-30 20:34:05 +08:00
|
|
|
printk(KERN_WARNING "fix_to_virt(FIX_BTMAP_END): %08lx\n",
|
2008-02-10 06:24:09 +08:00
|
|
|
fix_to_virt(FIX_BTMAP_END));
|
2008-01-30 20:34:05 +08:00
|
|
|
|
|
|
|
printk(KERN_WARNING "FIX_BTMAP_END: %d\n", FIX_BTMAP_END);
|
|
|
|
printk(KERN_WARNING "FIX_BTMAP_BEGIN: %d\n",
|
|
|
|
FIX_BTMAP_BEGIN);
|
2008-01-30 20:33:49 +08:00
|
|
|
}
|
2008-01-30 20:33:44 +08:00
|
|
|
}
|
|
|
|
|
2014-04-08 06:39:49 +08:00
|
|
|
void __init __early_set_fixmap(enum fixed_addresses idx,
|
|
|
|
phys_addr_t phys, pgprot_t flags)
|
2008-01-30 20:33:44 +08:00
|
|
|
{
|
2008-02-10 06:24:09 +08:00
|
|
|
unsigned long addr = __fix_to_virt(idx);
|
|
|
|
pte_t *pte;
|
2008-01-30 20:33:44 +08:00
|
|
|
|
|
|
|
if (idx >= __end_of_fixed_addresses) {
|
|
|
|
BUG();
|
|
|
|
return;
|
|
|
|
}
|
2008-01-30 20:33:44 +08:00
|
|
|
pte = early_ioremap_pte(addr);
|
2008-06-25 12:19:03 +08:00
|
|
|
|
2008-01-30 20:33:44 +08:00
|
|
|
if (pgprot_val(flags))
|
2008-02-10 06:24:09 +08:00
|
|
|
set_pte(pte, pfn_pte(phys >> PAGE_SHIFT, flags));
|
2008-01-30 20:33:44 +08:00
|
|
|
else
|
2008-06-25 12:19:19 +08:00
|
|
|
pte_clear(&init_mm, addr, pte);
|
2008-01-30 20:33:44 +08:00
|
|
|
__flush_tlb_one(addr);
|
|
|
|
}
|