[v3,0/7] hugetlb: parallelize hugetlb page init on boot

  Hi all, hugetlb init parallelization has now been updated to v3.

This series is tested on next-20240102 and can not be applied to v6.7-rc8.

Update Summary:
- Select CONFIG_PADATA as we use padata_do_multithreaded
- Fix a race condition in h->next_nid_to_alloc
- Fix local variable initialization issues
- Remove RFC tag

Thanks to the testing by David Rientjes, we now know that this patch reduce
hugetlb 1G initialization time from 77s to 18.3s on a 12T machine[4].

# Introduction
Hugetlb initialization during boot takes up a considerable amount of time.
For instance, on a 2TB system, initializing 1,800 1GB huge pages takes 1-2
seconds out of 10 seconds. Initializing 11,776 1GB pages on a 12TB Intel
host takes more than 1 minute[1]. This is a noteworthy figure.

Inspired by [2] and [3], hugetlb initialization can also be accelerated
through parallelization. Kernel already has infrastructure like
padata_do_multithreaded, this patch uses it to achieve effective results
by minimal modifications.

[1] https://lore.kernel.org/all/783f8bac-55b8-5b95-eb6a-11a583675000@google.com/
[2] https://lore.kernel.org/all/20200527173608.2885243-1-daniel.m.jordan@oracle.com/
[3] https://lore.kernel.org/all/20230906112605.2286994-1-usama.arif@bytedance.com/
[4] https://lore.kernel.org/all/76becfc1-e609-e3e8-2966-4053143170b6@google.com/

# Test result
        test          no patch(ms)   patched(ms)   saved   
 ------------------- -------------- ------------- -------- 
  256c2t(4 node) 1G           4745          2024   57.34%
  128c1t(2 node) 1G           3358          1712   49.02%
      12t        1G          77000         18300   76.23%

  256c2t(4 node) 2M           3336          1051   68.52%
  128c1t(2 node) 2M           1943           716   63.15%

# Change log
Changes in v3:
- Select CONFIG_PADATA as we use padata_do_multithreaded
- Fix a race condition in h->next_nid_to_alloc
- Fix local variable initialization issues
- Remove RFC tag

Changes in v2:
- https://lore.kernel.org/all/20231208025240.4744-1-gang.li@linux.dev/
- Reduce complexity with `padata_do_multithreaded`
- Support 1G hugetlb

v1:
- https://lore.kernel.org/all/20231123133036.68540-1-gang.li@linux.dev/
- parallelize 2M hugetlb initialization with workqueue

Gang Li (7):
  hugetlb: code clean for hugetlb_hstate_alloc_pages
  hugetlb: split hugetlb_hstate_alloc_pages
  padata: dispatch works on different nodes
  hugetlb: pass *next_nid_to_alloc directly to
    for_each_node_mask_to_alloc
  hugetlb: have CONFIG_HUGETLBFS select CONFIG_PADATA
  hugetlb: parallelize 2M hugetlb allocation and initialization
  hugetlb: parallelize 1G hugetlb initialization

 fs/Kconfig              |   1 +
 include/linux/hugetlb.h |   2 +-
 include/linux/padata.h  |   3 +
 kernel/padata.c         |   8 +-
 mm/hugetlb.c            | 224 +++++++++++++++++++++++++++-------------
 mm/mm_init.c            |   1 +
 6 files changed, 163 insertions(+), 76 deletions(-)
  

On Tue, 2 Jan 2024, Gang Li wrote:

> Hi all, hugetlb init parallelization has now been updated to v3.
> 
> This series is tested on next-20240102 and can not be applied to v6.7-rc8.
> 
> Update Summary:
> - Select CONFIG_PADATA as we use padata_do_multithreaded
> - Fix a race condition in h->next_nid_to_alloc
> - Fix local variable initialization issues
> - Remove RFC tag
> 
> Thanks to the testing by David Rientjes, we now know that this patch reduce
> hugetlb 1G initialization time from 77s to 18.3s on a 12T machine[4].
> 
> # Introduction
> Hugetlb initialization during boot takes up a considerable amount of time.
> For instance, on a 2TB system, initializing 1,800 1GB huge pages takes 1-2
> seconds out of 10 seconds. Initializing 11,776 1GB pages on a 12TB Intel
> host takes more than 1 minute[1]. This is a noteworthy figure.
> 
> Inspired by [2] and [3], hugetlb initialization can also be accelerated
> through parallelization. Kernel already has infrastructure like
> padata_do_multithreaded, this patch uses it to achieve effective results
> by minimal modifications.
> 
> [1] https://lore.kernel.org/all/783f8bac-55b8-5b95-eb6a-11a583675000@google.com/
> [2] https://lore.kernel.org/all/20200527173608.2885243-1-daniel.m.jordan@oracle.com/
> [3] https://lore.kernel.org/all/20230906112605.2286994-1-usama.arif@bytedance.com/
> [4] https://lore.kernel.org/all/76becfc1-e609-e3e8-2966-4053143170b6@google.com/
> 
> # Test result
>         test          no patch(ms)   patched(ms)   saved   
>  ------------------- -------------- ------------- -------- 
>   256c2t(4 node) 1G           4745          2024   57.34%
>   128c1t(2 node) 1G           3358          1712   49.02%
>       12t        1G          77000         18300   76.23%
> 
>   256c2t(4 node) 2M           3336          1051   68.52%
>   128c1t(2 node) 2M           1943           716   63.15%
> 

I tested 1GB hugetlb on a smaller AMD host with the following:

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3301,7 +3301,7 @@ int alloc_bootmem_huge_page(struct hstate *h, int nid)
 int __alloc_bootmem_huge_page(struct hstate *h, int nid)
 {
        struct huge_bootmem_page *m = NULL; /* initialize for clang */
-       int nr_nodes, node;
+       int nr_nodes, node = nid;
 
        /* do node specific alloc */
        if (nid != NUMA_NO_NODE) {

After the build error is fixed, feel free to add:

	Tested-by: David Rientjes <rientjes@google.com>

to each patch.  I think Andrew will probably take a build fix up as a
delta on top of patch 4 rather than sending a whole new series unless
there is other feedback that you receive.
  

On 2024/1/3 09:52, David Rientjes wrote:
> 
> I tested 1GB hugetlb on a smaller AMD host with the following:
> 
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -3301,7 +3301,7 @@ int alloc_bootmem_huge_page(struct hstate *h, int nid)
>   int __alloc_bootmem_huge_page(struct hstate *h, int nid)
>   {
>          struct huge_bootmem_page *m = NULL; /* initialize for clang */
> -       int nr_nodes, node;
> +       int nr_nodes, node = nid;
>   
>          /* do node specific alloc */
>          if (nid != NUMA_NO_NODE) {
> 

Oh, if nid != NUMA_NO_NODE and memblock_alloc_try_nid_raw succeed,
`node` must take the value of `nid`.

Otherwise, list_add(&m->list, &huge_boot_pages[node]) will not be
executed correctly.

> After the build error is fixed, feel free to add:
> 
> 	Tested-by: David Rientjes <rientjes@google.com>
> 

Thanks!

> to each patch.  I think Andrew will probably take a build fix up as a
> delta on top of patch 4 rather than sending a whole new series unless
> there is other feedback that you receive.