Revision 5b77e95dd7790ff6c8fbf1cd8d0104ebed818a03 authored by Alexander Potapenko on 02 April 2019, 11:28:13 UTC, committed by Ingo Molnar on 06 April 2019, 07:52:02 UTC
There's a number of problems with how arch/x86/include/asm/bitops.h is currently using assembly constraints for the memory region bitops are modifying: 1) Use memory clobber in bitops that touch arbitrary memory Certain bit operations that read/write bits take a base pointer and an arbitrarily large offset to address the bit relative to that base. Inline assembly constraints aren't expressive enough to tell the compiler that the assembly directive is going to touch a specific memory location of unknown size, therefore we have to use the "memory" clobber to indicate that the assembly is going to access memory locations other than those listed in the inputs/outputs. To indicate that BTR/BTS instructions don't necessarily touch the first sizeof(long) bytes of the argument, we also move the address to assembly inputs. This particular change leads to size increase of 124 kernel functions in a defconfig build. For some of them the diff is in NOP operations, other end up re-reading values from memory and may potentially slow down the execution. But without these clobbers the compiler is free to cache the contents of the bitmaps and use them as if they weren't changed by the inline assembly. 2) Use byte-sized arguments for operations touching single bytes. Passing a long value to ANDB/ORB/XORB instructions makes the compiler treat sizeof(long) bytes as being clobbered, which isn't the case. This may theoretically lead to worse code in the case of heavy optimization. Practical impact: I've built a defconfig kernel and looked through some of the functions generated by GCC 7.3.0 with and without this clobber, and didn't spot any miscompilations. However there is a (trivial) theoretical case where this code leads to miscompilation: https://lkml.org/lkml/2019/3/28/393 using just GCC 8.3.0 with -O2. It isn't hard to imagine someone writes such a function in the kernel someday. So the primary motivation is to fix an existing misuse of the asm directive, which happens to work in certain configurations now, but isn't guaranteed to work under different circumstances. [ --mingo: Added -stable tag because defconfig only builds a fraction of the kernel and the trivial testcase looks normal enough to be used in existing or in-development code. ] Signed-off-by: Alexander Potapenko <glider@google.com> Cc: <stable@vger.kernel.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: James Y Knight <jyknight@google.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20190402112813.193378-1-glider@google.com [ Edited the changelog, tidied up one of the defines. ] Signed-off-by: Ingo Molnar <mingo@kernel.org>
1 parent faa3604
io-mapping.txt
========================
The io_mapping functions
========================
API
===
The io_mapping functions in linux/io-mapping.h provide an abstraction for
efficiently mapping small regions of an I/O device to the CPU. The initial
usage is to support the large graphics aperture on 32-bit processors where
ioremap_wc cannot be used to statically map the entire aperture to the CPU
as it would consume too much of the kernel address space.
A mapping object is created during driver initialization using::
struct io_mapping *io_mapping_create_wc(unsigned long base,
unsigned long size)
'base' is the bus address of the region to be made
mappable, while 'size' indicates how large a mapping region to
enable. Both are in bytes.
This _wc variant provides a mapping which may only be used
with the io_mapping_map_atomic_wc or io_mapping_map_wc.
With this mapping object, individual pages can be mapped either atomically
or not, depending on the necessary scheduling environment. Of course, atomic
maps are more efficient::
void *io_mapping_map_atomic_wc(struct io_mapping *mapping,
unsigned long offset)
'offset' is the offset within the defined mapping region.
Accessing addresses beyond the region specified in the
creation function yields undefined results. Using an offset
which is not page aligned yields an undefined result. The
return value points to a single page in CPU address space.
This _wc variant returns a write-combining map to the
page and may only be used with mappings created by
io_mapping_create_wc
Note that the task may not sleep while holding this page
mapped.
::
void io_mapping_unmap_atomic(void *vaddr)
'vaddr' must be the value returned by the last
io_mapping_map_atomic_wc call. This unmaps the specified
page and allows the task to sleep once again.
If you need to sleep while holding the lock, you can use the non-atomic
variant, although they may be significantly slower.
::
void *io_mapping_map_wc(struct io_mapping *mapping,
unsigned long offset)
This works like io_mapping_map_atomic_wc except it allows
the task to sleep while holding the page mapped.
::
void io_mapping_unmap(void *vaddr)
This works like io_mapping_unmap_atomic, except it is used
for pages mapped with io_mapping_map_wc.
At driver close time, the io_mapping object must be freed::
void io_mapping_free(struct io_mapping *mapping)
Current Implementation
======================
The initial implementation of these functions uses existing mapping
mechanisms and so provides only an abstraction layer and no new
functionality.
On 64-bit processors, io_mapping_create_wc calls ioremap_wc for the whole
range, creating a permanent kernel-visible mapping to the resource. The
map_atomic and map functions add the requested offset to the base of the
virtual address returned by ioremap_wc.
On 32-bit processors with HIGHMEM defined, io_mapping_map_atomic_wc uses
kmap_atomic_pfn to map the specified page in an atomic fashion;
kmap_atomic_pfn isn't really supposed to be used with device pages, but it
provides an efficient mapping for this usage.
On 32-bit processors without HIGHMEM defined, io_mapping_map_atomic_wc and
io_mapping_map_wc both use ioremap_wc, a terribly inefficient function which
performs an IPI to inform all processors about the new mapping. This results
in a significant performance penalty.
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...