[Mesa-dev] [PATCH 1/9] util: Add a virtual memory allocator
Marek Olšák
maraeo at gmail.com
Wed May 2 19:05:57 UTC 2018
Hi Jason,
The radeon kernel driver only allows 4GB or 8GB of virtual address space,
so we are talking about 33 usable bits. The radeon winsys also has to
support 32-bit allocations where the high bits are 0, and it uses a
separate allocator if the address space size is 8GB, or not if it's just
4GB.
I you want to duplicate this, that's totally OK with me. I'd much rather
have the radeon winsys code duplicated than having it broken randomly by
changes to common code. Right now I know that if nobody is going to touch
it in the winsys, it's not going to break.
Marek
On Wed, May 2, 2018 at 2:36 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
> Marek & Nicolai,
>
> FYI, I am NOT trying to NIH anything here. My hope was that eventually,
> the radeon winsys code and the Intel drivers should share an allocator. I
> considered starting by pulling the one out of the radeon winsys code and
> modifying it but there were a couple of issues (mentioned below) and I
> didn't want to start stuff off by changing the behavior of your driver.
> :-) Please review and give your opinions as to whether or not the version
> here would work for radeon's use-cases.
>
> --Jason
>
> On Wed, May 2, 2018 at 9:01 AM, Scott D Phillips <
> scott.d.phillips at intel.com> wrote:
>
>> From: Jason Ekstrand <jason.ekstrand at intel.com>
>>
>> This is simple linear-walk first-fit allocator roughly based on the
>> allocator in the radeon winsys code. This allocator has two primary
>> functional differences:
>>
>> 1) It cleanly returns 0 on allocation failure
>>
>> 2) It allocates addresses top-down instead of bottom-up.
>>
>> The second one is needed for Intel because high addresses (with bit 47
>> set) need to be canonicalized in order to work properly. If we allocate
>> bottom-up, then high addresses will be very rare (if they ever happen).
>> We'd rather always have high addresses so that the canonicalization code
>> gets better testing.
>> ---
>> src/util/Makefile.sources | 4 +-
>> src/util/meson.build | 2 +
>> src/util/vma.c | 231 ++++++++++++++++++++++++++++++
>> ++++++++++++++++
>> src/util/vma.h | 53 +++++++++++
>> 4 files changed, 289 insertions(+), 1 deletion(-)
>> create mode 100644 src/util/vma.c
>> create mode 100644 src/util/vma.h
>>
>> diff --git a/src/util/Makefile.sources b/src/util/Makefile.sources
>> index 104ecae8ed3..534520ce763 100644
>> --- a/src/util/Makefile.sources
>> +++ b/src/util/Makefile.sources
>> @@ -56,7 +56,9 @@ MESA_UTIL_FILES := \
>> u_string.h \
>> u_thread.h \
>> u_vector.c \
>> - u_vector.h
>> + u_vector.h \
>> + vma.c \
>> + vma.h
>>
>> MESA_UTIL_GENERATED_FILES = \
>> format_srgb.c
>> diff --git a/src/util/meson.build b/src/util/meson.build
>> index eece1cefef6..14660e0fa0c 100644
>> --- a/src/util/meson.build
>> +++ b/src/util/meson.build
>> @@ -81,6 +81,8 @@ files_mesa_util = files(
>> 'u_thread.h',
>> 'u_vector.c',
>> 'u_vector.h',
>> + 'vma.c',
>> + 'vma.h',
>> )
>>
>> install_data('drirc', install_dir : get_option('sysconfdir'))
>> diff --git a/src/util/vma.c b/src/util/vma.c
>> new file mode 100644
>> index 00000000000..0d4e097e21f
>> --- /dev/null
>> +++ b/src/util/vma.c
>> @@ -0,0 +1,231 @@
>> +/*
>> + * Copyright © 2018 Intel Corporation
>> + *
>> + * 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 (including the
>> next
>> + * paragraph) 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
>> + * THE AUTHORS OR COPYRIGHT HOLDERS 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.
>> + */
>> +
>> +#include <stdlib.h>
>> +
>> +#include "util/u_math.h"
>> +#include "util/vma.h"
>> +
>> +struct util_vma_hole {
>> + struct list_head link;
>> + uint64_t offset;
>> + uint64_t size;
>> +};
>> +
>> +#define util_vma_foreach_hole(_hole, _heap) \
>> + list_for_each_entry(struct util_vma_hole, _hole, &(_heap)->holes,
>> link)
>> +
>> +#define util_vma_foreach_hole_safe(_hole, _heap) \
>> + list_for_each_entry_safe(struct util_vma_hole, _hole,
>> &(_heap)->holes, link)
>> +
>> +void
>> +util_vma_heap_init(struct util_vma_heap *heap,
>> + uint64_t start, uint64_t size)
>> +{
>> + list_inithead(&heap->holes);
>> + util_vma_heap_free(heap, start, size);
>> +}
>> +
>> +void
>> +util_vma_heap_finish(struct util_vma_heap *heap)
>> +{
>> + util_vma_foreach_hole_safe(hole, heap)
>> + free(hole);
>> +}
>> +
>> +static void
>> +util_vma_heap_validate(struct util_vma_heap *heap)
>> +{
>> + uint64_t prev_offset = 0;
>> + util_vma_foreach_hole(hole, heap) {
>> + assert(hole->offset > 0);
>> + assert(hole->size > 0);
>> +
>> + if (&hole->link == heap->holes.next) {
>> + /* This must be the top-most hole. Assert that, if it
>> overflows, it
>> + * overflows to 0, i.e. 2^64.
>> + */
>> + assert(hole->size + hole->offset == 0 ||
>> + hole->size + hole->offset > hole->offset);
>> + } else {
>> + /* This is not the top-most hole so it must not overflow and, in
>> + * fact, must be strictly lower than the top-most hole. If
>> + * hole->size + hole->offset == prev_offset, then we failed to
>> join
>> + * holes during a util_vma_heap_free.
>> + */
>> + assert(hole->size + hole->offset > hole->offset &&
>> + hole->size + hole->offset < prev_offset);
>> + }
>> + prev_offset = hole->offset;
>> + }
>> +}
>> +
>> +uint64_t
>> +util_vma_heap_alloc(struct util_vma_heap *heap,
>> + uint64_t size, uint64_t alignment)
>> +{
>> + /* The caller is expected to reject zero-size allocations */
>> + assert(size > 0);
>> +
>> + assert(util_is_power_of_two_nonzero(alignment));
>> +
>> + util_vma_heap_validate(heap);
>> +
>> + util_vma_foreach_hole_safe(hole, heap) {
>> + if (size > hole->size)
>> + continue;
>> +
>> + /* Compute the offset as the highest address where a chunk of the
>> given
>> + * size can be without going over the top of the hole.
>> + *
>> + * This calculation is known to not overflow because we know that
>> + * hole->size + hole->offset can only overflow to 0 and size > 0.
>> + */
>> + uint64_t offset = (hole->size - size) + hole->offset;
>> +
>> + /* Align the offset. We align down and not up because we are
>> allocating
>> + * from the top of the hole and not the bottom.
>> + */
>> + offset &= ~(alignment - 1);
>> +
>> + if (offset < hole->offset)
>> + continue;
>> +
>> + if (offset == hole->offset && size == hole->size) {
>> + /* Just get rid of the hole. */
>> + list_del(&hole->link);
>> + free(hole);
>> + util_vma_heap_validate(heap);
>> + return offset;
>> + }
>> +
>> + assert(offset - hole->offset <= hole->size - size);
>> + uint64_t waste = (hole->size - size) - (offset - hole->offset);
>> + if (waste == 0) {
>> + /* We allocated at the top. Shrink the hole down. */
>> + hole->size -= size;
>> + util_vma_heap_validate(heap);
>> + return offset;
>> + }
>> +
>> + if (offset == hole->offset) {
>> + /* We allocated at the bottom. Shrink the hole up. */
>> + hole->offset += size;
>> + hole->size -= size;
>> + util_vma_heap_validate(heap);
>> + return offset;
>> + }
>> +
>> + /* We allocated in the middle. We need to split the old hole into
>> two
>> + * holes, one high and one low.
>> + */
>> + struct util_vma_hole *high_hole = calloc(1, sizeof(*hole));
>> + high_hole->offset = offset + size;
>> + high_hole->size = waste;
>> +
>> + /* Adjust the hole to be the amount of space left at he bottom of
>> the
>> + * original hole.
>> + */
>> + hole->size = offset - hole->offset;
>> +
>> + /* Place the new hole before the old hole so that the list is in
>> order
>> + * from high to low.
>> + */
>> + list_addtail(&high_hole->link, &hole->link);
>> +
>> + util_vma_heap_validate(heap);
>> +
>> + return offset;
>> + }
>> +
>> + /* Failed to allocate */
>> + return 0;
>> +}
>> +
>> +void
>> +util_vma_heap_free(struct util_vma_heap *heap,
>> + uint64_t offset, uint64_t size)
>> +{
>> + /* An offset of 0 is reserved for allocation failure. It is not a
>> valid
>> + * address and cannot be freed.
>> + */
>> + assert(offset > 0);
>> +
>> + /* Freeing something with a size of 0 is also not valid. */
>> + assert(size > 0);
>> +
>> + /* It's possible for offset + size to wrap around if we touch the top
>> of
>> + * the 64-bit address space, but we cannot go any higher than 2^64.
>> + */
>> + assert(offset + size == 0 || offset + size > offset);
>> +
>> + util_vma_heap_validate(heap);
>> +
>> + /* Find immediately higher and lower holes if they exist. */
>> + struct util_vma_hole *high_hole = NULL, *low_hole = NULL;
>> + util_vma_foreach_hole(hole, heap) {
>> + if (hole->offset <= offset) {
>> + low_hole = hole;
>> + break;
>> + }
>> + high_hole = hole;
>> + }
>> +
>> + if (high_hole)
>> + assert(offset + size <= high_hole->offset);
>> + bool high_adjacent = high_hole && offset + size == high_hole->offset;
>> +
>> + if (low_hole) {
>> + assert(low_hole->offset + low_hole->size > low_hole->offset);
>> + assert(low_hole->offset + low_hole->size <= offset);
>> + }
>> + bool low_adjacent = low_hole && low_hole->offset + low_hole->size ==
>> offset;
>> +
>> + if (low_adjacent && high_adjacent) {
>> + /* Merge the two holes */
>> + low_hole->size += size + high_hole->size;
>> + list_del(&high_hole->link);
>> + free(high_hole);
>> + } else if (low_adjacent) {
>> + /* Merge into the low hole */
>> + low_hole->size += size;
>> + } else if (high_adjacent) {
>> + /* Merge into the high hole */
>> + high_hole->offset = offset;
>> + high_hole->size += size;
>> + } else {
>> + /* Neither hole is adjacent; make a new one */
>> + struct util_vma_hole *hole = calloc(1, sizeof(*hole));
>> +
>> + hole->offset = offset;
>> + hole->size = size;
>> +
>> + /* Add it after the high hole so we maintain high-to-low ordering
>> */
>> + if (high_hole)
>> + list_add(&hole->link, &high_hole->link);
>> + else
>> + list_add(&hole->link, &heap->holes);
>> + }
>> +
>> + util_vma_heap_validate(heap);
>> +}
>> diff --git a/src/util/vma.h b/src/util/vma.h
>> new file mode 100644
>> index 00000000000..ed69914e4cb
>> --- /dev/null
>> +++ b/src/util/vma.h
>> @@ -0,0 +1,53 @@
>> +/*
>> + * Copyright © 2018 Intel Corporation
>> + *
>> + * 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 (including the
>> next
>> + * paragraph) 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
>> + * THE AUTHORS OR COPYRIGHT HOLDERS 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.
>> + */
>> +
>> +#ifndef _UTIL_VMA_H
>> +#define _UTIL_VMA_H
>> +
>> +#include <stdint.h>
>> +
>> +#include "list.h"
>> +
>> +#ifdef __cplusplus
>> +extern "C" {
>> +#endif
>> +
>> +struct util_vma_heap {
>> + struct list_head holes;
>> +};
>> +
>> +void util_vma_heap_init(struct util_vma_heap *heap,
>> + uint64_t start, uint64_t size);
>> +void util_vma_heap_finish(struct util_vma_heap *heap);
>> +
>> +uint64_t util_vma_heap_alloc(struct util_vma_heap *heap,
>> + uint64_t size, uint64_t alignment);
>> +
>> +void util_vma_heap_free(struct util_vma_heap *heap,
>> + uint64_t offset, uint64_t size);
>> +
>> +#ifdef __cplusplus
>> +} /* extern C */
>> +#endif
>> +
>> +#endif /* _UTIL_DEBUG_H */
>> --
>> 2.14.3
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180502/a4c06a3e/attachment-0001.html>
More information about the mesa-dev
mailing list