memblock: Introduce memblock_reserve_node()

Message ID 20230624024622.2959376-1-yajun.deng@linux.dev
State New
Headers
Series memblock: Introduce memblock_reserve_node() |

Commit Message

Yajun Deng June 24, 2023, 2:46 a.m. UTC
  It only returns address now in memblock_find_in_range_node(), we can add a
parameter pointing to integer for node id of the range, which can be used
to pass the node id to the new reserve region.

Introduce memblock_reserve_node() so that the node id can be passed to
the reserve region in memblock_alloc_range_nid().

Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
---
 mm/memblock.c | 39 +++++++++++++++++++++++++++------------
 1 file changed, 27 insertions(+), 12 deletions(-)
  

Comments

Mike Rapoport June 25, 2023, 5:08 a.m. UTC | #1
On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
> It only returns address now in memblock_find_in_range_node(), we can add a
> parameter pointing to integer for node id of the range, which can be used
> to pass the node id to the new reserve region.
> 
> Introduce memblock_reserve_node() so that the node id can be passed to
> the reserve region in memblock_alloc_range_nid().
> 
> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>

What problem does this patch solve?
  
Yajun Deng June 25, 2023, 7:39 a.m. UTC | #2
June 25, 2023 1:08 PM, "Mike Rapoport" <rppt@kernel.org> wrote:

> On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
> 
>> It only returns address now in memblock_find_in_range_node(), we can add a
>> parameter pointing to integer for node id of the range, which can be used
>> to pass the node id to the new reserve region.
>> 
>> Introduce memblock_reserve_node() so that the node id can be passed to
>> the reserve region in memblock_alloc_range_nid().
>> 
>> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
> 
> What problem does this patch solve?
>

If we set nid and flags in memblock_alloc_range_nid(), we may not need
memblock_set_node() in memmap_init_reserved_pages().

I tested this patch and delete memblock_set_node() in memmap_init_reserved_pages().
It works fine. I did not delete memblock_set_node() in this patch just in case.
 
> --
> Sincerely yours,
> Mike.
  
Mike Rapoport June 26, 2023, 6:21 a.m. UTC | #3
On Sun, Jun 25, 2023 at 07:39:10AM +0000, Yajun Deng wrote:
> June 25, 2023 1:08 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
> 
> > On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
> > 
> >> It only returns address now in memblock_find_in_range_node(), we can add a
> >> parameter pointing to integer for node id of the range, which can be used
> >> to pass the node id to the new reserve region.
> >> 
> >> Introduce memblock_reserve_node() so that the node id can be passed to
> >> the reserve region in memblock_alloc_range_nid().
> >> 
> >> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
> > 
> > What problem does this patch solve?
> >
> 
> If we set nid and flags in memblock_alloc_range_nid(), we may not need
> memblock_set_node() in memmap_init_reserved_pages().

When memblock_reserve() is called before NUMA setup, the node ids are still
unset in memblock.memory, so very early reservations will be missed and we
still have to update node ids in memblock.reserved later.
 
> I tested this patch and delete memblock_set_node() in memmap_init_reserved_pages().
> It works fine. I did not delete memblock_set_node() in this patch just in case.
  
Yajun Deng June 27, 2023, 12:13 a.m. UTC | #4
June 26, 2023 2:21 PM, "Mike Rapoport" <rppt@kernel.org> wrote:

> On Sun, Jun 25, 2023 at 07:39:10AM +0000, Yajun Deng wrote:
> 
>> June 25, 2023 1:08 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
>> 
>> On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
>> 
>> It only returns address now in memblock_find_in_range_node(), we can add a
>> parameter pointing to integer for node id of the range, which can be used
>> to pass the node id to the new reserve region.
>> 
>> Introduce memblock_reserve_node() so that the node id can be passed to
>> the reserve region in memblock_alloc_range_nid().
>> 
>> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
>> 
>> What problem does this patch solve?
>> 
>> If we set nid and flags in memblock_alloc_range_nid(), we may not need
>> memblock_set_node() in memmap_init_reserved_pages().
> 
> When memblock_reserve() is called before NUMA setup, the node ids are still
> unset in memblock.memory, so very early reservations will be missed and we
> still have to update node ids in memblock.reserved later.

Even so, we still need to pass the 'flags' to the new reserve region.
choose_memblock_flags() may return MEMBLOCK_MIRROR in memblock_alloc_range_nid(),
memblock_reserve() couldn't pass this flag in this case.


>> I tested this patch and delete memblock_set_node() in memmap_init_reserved_pages().
>> It works fine. I did not delete memblock_set_node() in this patch just in case.
> 
> --
> Sincerely yours,
> Mike.
  
Mike Rapoport June 27, 2023, 2:33 p.m. UTC | #5
On Tue, Jun 27, 2023 at 12:13:16AM +0000, Yajun Deng wrote:
> June 26, 2023 2:21 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
> 
> > On Sun, Jun 25, 2023 at 07:39:10AM +0000, Yajun Deng wrote:
> > 
> >> June 25, 2023 1:08 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
> >> 
> >> On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
> >> 
> >> It only returns address now in memblock_find_in_range_node(), we can add a
> >> parameter pointing to integer for node id of the range, which can be used
> >> to pass the node id to the new reserve region.
> >> 
> >> Introduce memblock_reserve_node() so that the node id can be passed to
> >> the reserve region in memblock_alloc_range_nid().
> >> 
> >> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
> >> 
> >> What problem does this patch solve?
> >> 
> >> If we set nid and flags in memblock_alloc_range_nid(), we may not need
> >> memblock_set_node() in memmap_init_reserved_pages().
> > 
> > When memblock_reserve() is called before NUMA setup, the node ids are still
> > unset in memblock.memory, so very early reservations will be missed and we
> > still have to update node ids in memblock.reserved later.
> 
> Even so, we still need to pass the 'flags' to the new reserve region.
> choose_memblock_flags() may return MEMBLOCK_MIRROR in memblock_alloc_range_nid(),
> memblock_reserve() couldn't pass this flag in this case.
 
flags are only relevant to memblock.memory, we don't care about the flags
in memblock.reserved.
 
> >> I tested this patch and delete memblock_set_node() in memmap_init_reserved_pages().
> >> It works fine. I did not delete memblock_set_node() in this patch just in case.
> > 
> > --
> > Sincerely yours,
> > Mike.
  
Yajun Deng June 28, 2023, 1:49 a.m. UTC | #6
June 27, 2023 10:33 PM, "Mike Rapoport" <rppt@kernel.org> wrote:

> On Tue, Jun 27, 2023 at 12:13:16AM +0000, Yajun Deng wrote:
> 
>> June 26, 2023 2:21 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
>> 
>> On Sun, Jun 25, 2023 at 07:39:10AM +0000, Yajun Deng wrote:
>> 
>> June 25, 2023 1:08 PM, "Mike Rapoport" <rppt@kernel.org> wrote:
>> 
>> On Sat, Jun 24, 2023 at 10:46:22AM +0800, Yajun Deng wrote:
>> 
>> It only returns address now in memblock_find_in_range_node(), we can add a
>> parameter pointing to integer for node id of the range, which can be used
>> to pass the node id to the new reserve region.
>> 
>> Introduce memblock_reserve_node() so that the node id can be passed to
>> the reserve region in memblock_alloc_range_nid().
>> 
>> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
>> 
>> What problem does this patch solve?
>> 
>> If we set nid and flags in memblock_alloc_range_nid(), we may not need
>> memblock_set_node() in memmap_init_reserved_pages().
>> 
>> When memblock_reserve() is called before NUMA setup, the node ids are still
>> unset in memblock.memory, so very early reservations will be missed and we
>> still have to update node ids in memblock.reserved later.
>> 
>> Even so, we still need to pass the 'flags' to the new reserve region.
>> choose_memblock_flags() may return MEMBLOCK_MIRROR in memblock_alloc_range_nid(),
>> memblock_reserve() couldn't pass this flag in this case.
> 
> flags are only relevant to memblock.memory, we don't care about the flags
> in memblock.reserved.
> 

get it.

>> I tested this patch and delete memblock_set_node() in memmap_init_reserved_pages().
>> It works fine. I did not delete memblock_set_node() in this patch just in case.
>> 
>> --
>> Sincerely yours,
>> Mike.
> 
> --
> Sincerely yours,
> Mike.
  

Patch

diff --git a/mm/memblock.c b/mm/memblock.c
index f9e61e565a53..6b5f6c246458 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -204,6 +204,7 @@  bool __init_memblock memblock_overlaps_region(struct memblock_type *type,
  * @align: alignment of free area to find
  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
  * @flags: pick from blocks based on memory attributes
+ * @p_nid: ptr to int for nid of the range, can be %NULL
  *
  * Utility called from memblock_find_in_range_node(), find free area bottom-up.
  *
@@ -213,12 +214,12 @@  bool __init_memblock memblock_overlaps_region(struct memblock_type *type,
 static phys_addr_t __init_memblock
 __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
 				phys_addr_t size, phys_addr_t align, int nid,
-				enum memblock_flags flags)
+				enum memblock_flags flags, int *p_nid)
 {
 	phys_addr_t this_start, this_end, cand;
 	u64 i;
 
-	for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) {
+	for_each_free_mem_range(i, nid, flags, &this_start, &this_end, p_nid) {
 		this_start = clamp(this_start, start, end);
 		this_end = clamp(this_end, start, end);
 
@@ -239,6 +240,7 @@  __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
  * @align: alignment of free area to find
  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
  * @flags: pick from blocks based on memory attributes
+ * @p_nid: ptr to int for nid of the range, can be %NULL
  *
  * Utility called from memblock_find_in_range_node(), find free area top-down.
  *
@@ -248,13 +250,13 @@  __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
 static phys_addr_t __init_memblock
 __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
 			       phys_addr_t size, phys_addr_t align, int nid,
-			       enum memblock_flags flags)
+			       enum memblock_flags flags, int *p_nid)
 {
 	phys_addr_t this_start, this_end, cand;
 	u64 i;
 
 	for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end,
-					NULL) {
+					p_nid) {
 		this_start = clamp(this_start, start, end);
 		this_end = clamp(this_end, start, end);
 
@@ -278,6 +280,7 @@  __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
  *       %MEMBLOCK_ALLOC_ACCESSIBLE
  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
  * @flags: pick from blocks based on memory attributes
+ * @p_nid: ptr to int for nid of the range, can be %NULL
  *
  * Find @size free area aligned to @align in the specified range and node.
  *
@@ -287,7 +290,7 @@  __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
 static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
 					phys_addr_t align, phys_addr_t start,
 					phys_addr_t end, int nid,
-					enum memblock_flags flags)
+					enum memblock_flags flags, int *p_nid)
 {
 	/* pump up @end */
 	if (end == MEMBLOCK_ALLOC_ACCESSIBLE ||
@@ -300,10 +303,10 @@  static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
 
 	if (memblock_bottom_up())
 		return __memblock_find_range_bottom_up(start, end, size, align,
-						       nid, flags);
+						       nid, flags, p_nid);
 	else
 		return __memblock_find_range_top_down(start, end, size, align,
-						      nid, flags);
+						      nid, flags, p_nid);
 }
 
 /**
@@ -328,7 +331,7 @@  static phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
 
 again:
 	ret = memblock_find_in_range_node(size, align, start, end,
-					    NUMA_NO_NODE, flags);
+					    NUMA_NO_NODE, flags, NULL);
 
 	if (!ret && (flags & MEMBLOCK_MIRROR)) {
 		pr_warn_ratelimited("Could not allocate %pap bytes of mirrored memory\n",
@@ -863,6 +866,17 @@  int __init_memblock memblock_phys_free(phys_addr_t base, phys_addr_t size)
 	return memblock_remove_range(&memblock.reserved, base, size);
 }
 
+static int __init_memblock memblock_reserve_node(phys_addr_t base, phys_addr_t size,
+						 int nid, enum memblock_flags flags)
+{
+	phys_addr_t end = base + size - 1;
+
+	memblock_dbg("%s: [%pa-%pa] nid=%d flags=%x %pS\n", __func__,
+		     &base, &end, nid, flags, (void *)_RET_IP_);
+
+	return memblock_add_range(&memblock.reserved, base, size, nid, flags);
+}
+
 int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
 {
 	phys_addr_t end = base + size - 1;
@@ -1389,6 +1403,7 @@  phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
 {
 	enum memblock_flags flags = choose_memblock_flags();
 	phys_addr_t found;
+	int p_nid;
 
 	if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
 		nid = NUMA_NO_NODE;
@@ -1401,15 +1416,15 @@  phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
 
 again:
 	found = memblock_find_in_range_node(size, align, start, end, nid,
-					    flags);
-	if (found && !memblock_reserve(found, size))
+					    flags, &p_nid);
+	if (found && !memblock_reserve_node(found, size, p_nid, flags))
 		goto done;
 
 	if (nid != NUMA_NO_NODE && !exact_nid) {
 		found = memblock_find_in_range_node(size, align, start,
 						    end, NUMA_NO_NODE,
-						    flags);
-		if (found && !memblock_reserve(found, size))
+						    flags, &p_nid);
+		if (found && !memblock_reserve_node(found, size, p_nid, flags))
 			goto done;
 	}