The patch titled
Subject: mm,tmpfs: consider end of file write in shmem_is_huge
has been added to the -mm mm-unstable branch. Its filename is
mmtmpfs-consider-end-of-file-write-in-shmem_is_huge.patch
This patch will shortly appear at
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/mmtmpfs-consider-end-of-file-write-in-shmem_is_huge.patch
This patch will later appear in the mm-unstable branch at
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next via the mm-everything
branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there every 2-3 working days
------------------------------------------------------
From: Rik van Riel <riel@xxxxxxxxxxx>
Subject: mm,tmpfs: consider end of file write in shmem_is_huge
Date: Tue, 3 Sep 2024 11:19:28 -0400
Take the end of a file write into consideration when deciding whether or
not to use huge pages for tmpfs files when the tmpfs filesystem is mounted
with huge=within_size
This allows large writes that append to the end of a file to automatically
use large pages.
Doing 4MB sequential writes without fallocate to a 16GB tmpfs file with
fio. The numbers without THP or with huge=always stay the same, but the
performance with huge=within_size now matches that of huge=always.
huge before after
4kB pages 1560 MB/s 1560 MB/s
within_size 1560 MB/s 4720 MB/s
always: 4720 MB/s 4720 MB/s
Link: https://lkml.kernel.org/r/20240903111928.7171e60c@xxxxxxxxxxxxxxxxxxxx
Signed-off-by: Rik van Riel <riel@xxxxxxxxxxx>
Cc: Baolin Wang <baolin.wang@xxxxxxxxxxxxxxxxx>
Cc: Darrick J. Wong <djwong@xxxxxxxxxx>
Cc: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/xfs/scrub/xfile.c | 6 +--
fs/xfs/xfs_buf_mem.c | 2 -
include/linux/shmem_fs.h | 8 ++---
mm/huge_memory.c | 2 -
mm/khugepaged.c | 2 -
mm/shmem.c | 59 +++++++++++++++++++------------------
mm/userfaultfd.c | 2 -
7 files changed, 42 insertions(+), 39 deletions(-)
--- a/fs/xfs/scrub/xfile.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/fs/xfs/scrub/xfile.c
@@ -126,7 +126,7 @@ xfile_load(
unsigned int len;
unsigned int offset;
- if (shmem_get_folio(inode, pos >> PAGE_SHIFT, &folio,
+ if (shmem_get_folio(inode, pos >> PAGE_SHIFT, 0, &folio,
SGP_READ) < 0)
break;
if (!folio) {
@@ -196,7 +196,7 @@ xfile_store(
unsigned int len;
unsigned int offset;
- if (shmem_get_folio(inode, pos >> PAGE_SHIFT, &folio,
+ if (shmem_get_folio(inode, pos >> PAGE_SHIFT, 0, &folio,
SGP_CACHE) < 0)
break;
if (filemap_check_wb_err(inode->i_mapping, 0)) {
@@ -267,7 +267,7 @@ xfile_get_folio(
i_size_write(inode, pos + len);
pflags = memalloc_nofs_save();
- error = shmem_get_folio(inode, pos >> PAGE_SHIFT, &folio,
+ error = shmem_get_folio(inode, pos >> PAGE_SHIFT, 0, &folio,
(flags & XFILE_ALLOC) ? SGP_CACHE : SGP_READ);
memalloc_nofs_restore(pflags);
if (error)
--- a/fs/xfs/xfs_buf_mem.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/fs/xfs/xfs_buf_mem.c
@@ -149,7 +149,7 @@ xmbuf_map_page(
return -ENOMEM;
}
- error = shmem_get_folio(inode, pos >> PAGE_SHIFT, &folio, SGP_CACHE);
+ error = shmem_get_folio(inode, pos >> PAGE_SHIFT, 0, &folio, SGP_CACHE);
if (error)
return error;
--- a/include/linux/shmem_fs.h~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/include/linux/shmem_fs.h
@@ -113,11 +113,11 @@ int shmem_unuse(unsigned int type);
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
unsigned long shmem_allowable_huge_orders(struct inode *inode,
struct vm_area_struct *vma, pgoff_t index,
- bool shmem_huge_force);
+ loff_t write_end, bool shmem_huge_force);
#else
static inline unsigned long shmem_allowable_huge_orders(struct inode *inode,
struct vm_area_struct *vma, pgoff_t index,
- bool shmem_huge_force)
+ loff_t write_end, bool shmem_huge_force)
{
return 0;
}
@@ -143,8 +143,8 @@ enum sgp_type {
SGP_FALLOC, /* like SGP_WRITE, but make existing page Uptodate */
};
-int shmem_get_folio(struct inode *inode, pgoff_t index, struct folio **foliop,
- enum sgp_type sgp);
+int shmem_get_folio(struct inode *inode, pgoff_t index, loff_t write_end,
+ struct folio **foliop, enum sgp_type sgp);
struct folio *shmem_read_folio_gfp(struct address_space *mapping,
pgoff_t index, gfp_t gfp);
--- a/mm/huge_memory.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/mm/huge_memory.c
@@ -164,7 +164,7 @@ unsigned long __thp_vma_allowable_orders
*/
if (!in_pf && shmem_file(vma->vm_file))
return shmem_allowable_huge_orders(file_inode(vma->vm_file),
- vma, vma->vm_pgoff,
+ vma, vma->vm_pgoff, 0,
!enforce_sysfs);
if (!vma_is_anonymous(vma)) {
--- a/mm/khugepaged.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/mm/khugepaged.c
@@ -1870,7 +1870,7 @@ static int collapse_file(struct mm_struc
if (xa_is_value(folio) || !folio_test_uptodate(folio)) {
xas_unlock_irq(&xas);
/* swap in or instantiate fallocated page */
- if (shmem_get_folio(mapping->host, index,
+ if (shmem_get_folio(mapping->host, index, 0,
&folio, SGP_NOALLOC)) {
result = SCAN_FAIL;
goto xa_unlocked;
--- a/mm/shmem.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/mm/shmem.c
@@ -549,7 +549,8 @@ static bool shmem_confirm_swap(struct ad
static int shmem_huge __read_mostly = SHMEM_HUGE_NEVER;
static bool __shmem_huge_global_enabled(struct inode *inode, pgoff_t index,
- bool shmem_huge_force, struct vm_area_struct *vma,
+ loff_t write_end, bool shmem_huge_force,
+ struct vm_area_struct *vma,
unsigned long vm_flags)
{
struct mm_struct *mm = vma ? vma->vm_mm : NULL;
@@ -569,7 +570,8 @@ static bool __shmem_huge_global_enabled(
return true;
case SHMEM_HUGE_WITHIN_SIZE:
index = round_up(index + 1, HPAGE_PMD_NR);
- i_size = round_up(i_size_read(inode), PAGE_SIZE);
+ i_size = max(write_end, i_size_read(inode));
+ i_size = round_up(i_size, PAGE_SIZE);
if (i_size >> PAGE_SHIFT >= index)
return true;
fallthrough;
@@ -583,14 +585,14 @@ static bool __shmem_huge_global_enabled(
}
static bool shmem_huge_global_enabled(struct inode *inode, pgoff_t index,
- bool shmem_huge_force, struct vm_area_struct *vma,
- unsigned long vm_flags)
+ loff_t write_end, bool shmem_huge_force,
+ struct vm_area_struct *vma, unsigned long vm_flags)
{
if (HPAGE_PMD_ORDER > MAX_PAGECACHE_ORDER)
return false;
- return __shmem_huge_global_enabled(inode, index, shmem_huge_force,
- vma, vm_flags);
+ return __shmem_huge_global_enabled(inode, index, write_end,
+ shmem_huge_force, vma, vm_flags);
}
#if defined(CONFIG_SYSFS)
@@ -770,8 +772,8 @@ static unsigned long shmem_unused_huge_s
}
static bool shmem_huge_global_enabled(struct inode *inode, pgoff_t index,
- bool shmem_huge_force, struct vm_area_struct *vma,
- unsigned long vm_flags)
+ loff_t write_end, bool shmem_huge_force,
+ struct vm_area_struct *vma, unsigned long vm_flags)
{
return false;
}
@@ -978,7 +980,7 @@ static struct folio *shmem_get_partial_f
* (although in some cases this is just a waste of time).
*/
folio = NULL;
- shmem_get_folio(inode, index, &folio, SGP_READ);
+ shmem_get_folio(inode, index, 0, &folio, SGP_READ);
return folio;
}
@@ -1166,7 +1168,7 @@ static int shmem_getattr(struct mnt_idma
STATX_ATTR_NODUMP);
generic_fillattr(idmap, request_mask, inode, stat);
- if (shmem_huge_global_enabled(inode, 0, false, NULL, 0))
+ if (shmem_huge_global_enabled(inode, 0, 0, false, NULL, 0))
stat->blksize = HPAGE_PMD_SIZE;
if (request_mask & STATX_BTIME) {
@@ -1653,7 +1655,7 @@ static gfp_t limit_gfp_mask(gfp_t huge_g
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
unsigned long shmem_allowable_huge_orders(struct inode *inode,
struct vm_area_struct *vma, pgoff_t index,
- bool shmem_huge_force)
+ loff_t write_end, bool shmem_huge_force)
{
unsigned long mask = READ_ONCE(huge_shmem_orders_always);
unsigned long within_size_orders = READ_ONCE(huge_shmem_orders_within_size);
@@ -1670,8 +1672,8 @@ unsigned long shmem_allowable_huge_order
if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
return 0;
- global_huge = shmem_huge_global_enabled(inode, index, shmem_huge_force,
- vma, vm_flags);
+ global_huge = shmem_huge_global_enabled(inode, index, write_end,
+ shmem_huge_force, vma, vm_flags);
if (!vma || !vma_is_anon_shmem(vma)) {
/*
* For tmpfs, we now only support PMD sized THP if huge page
@@ -2231,8 +2233,8 @@ unlock:
* vmf and fault_type are only supplied by shmem_fault: otherwise they are NULL.
*/
static int shmem_get_folio_gfp(struct inode *inode, pgoff_t index,
- struct folio **foliop, enum sgp_type sgp, gfp_t gfp,
- struct vm_fault *vmf, vm_fault_t *fault_type)
+ loff_t write_end, struct folio **foliop, enum sgp_type sgp,
+ gfp_t gfp, struct vm_fault *vmf, vm_fault_t *fault_type)
{
struct vm_area_struct *vma = vmf ? vmf->vma : NULL;
struct mm_struct *fault_mm;
@@ -2312,7 +2314,7 @@ repeat:
}
/* Find hugepage orders that are allowed for anonymous shmem and tmpfs. */
- orders = shmem_allowable_huge_orders(inode, vma, index, false);
+ orders = shmem_allowable_huge_orders(inode, vma, index, write_end, false);
if (orders > 0) {
gfp_t huge_gfp;
@@ -2413,6 +2415,7 @@ unlock:
* shmem_get_folio - find, and lock a shmem folio.
* @inode: inode to search
* @index: the page index.
+ * @write_end: end of a write, could extend inode size
* @foliop: pointer to the folio if found
* @sgp: SGP_* flags to control behavior
*
@@ -2432,10 +2435,10 @@ unlock:
* Context: May sleep.
* Return: 0 if successful, else a negative error code.
*/
-int shmem_get_folio(struct inode *inode, pgoff_t index, struct folio **foliop,
- enum sgp_type sgp)
+int shmem_get_folio(struct inode *inode, pgoff_t index, loff_t write_end,
+ struct folio **foliop, enum sgp_type sgp)
{
- return shmem_get_folio_gfp(inode, index, foliop, sgp,
+ return shmem_get_folio_gfp(inode, index, write_end, foliop, sgp,
mapping_gfp_mask(inode->i_mapping), NULL, NULL);
}
EXPORT_SYMBOL_GPL(shmem_get_folio);
@@ -2530,7 +2533,7 @@ static vm_fault_t shmem_fault(struct vm_
}
WARN_ON_ONCE(vmf->page != NULL);
- err = shmem_get_folio_gfp(inode, vmf->pgoff, &folio, SGP_CACHE,
+ err = shmem_get_folio_gfp(inode, vmf->pgoff, 0, &folio, SGP_CACHE,
gfp, vmf, &ret);
if (err)
return vmf_error(err);
@@ -3040,7 +3043,7 @@ shmem_write_begin(struct file *file, str
return -EPERM;
}
- ret = shmem_get_folio(inode, index, &folio, SGP_WRITE);
+ ret = shmem_get_folio(inode, index, pos + len, &folio, SGP_WRITE);
if (ret)
return ret;
@@ -3111,7 +3114,7 @@ static ssize_t shmem_file_read_iter(stru
break;
}
- error = shmem_get_folio(inode, index, &folio, SGP_READ);
+ error = shmem_get_folio(inode, index, 0, &folio, SGP_READ);
if (error) {
if (error == -EINVAL)
error = 0;
@@ -3287,7 +3290,7 @@ static ssize_t shmem_file_splice_read(st
if (*ppos >= i_size_read(inode))
break;
- error = shmem_get_folio(inode, *ppos / PAGE_SIZE, &folio,
+ error = shmem_get_folio(inode, *ppos / PAGE_SIZE, 0, &folio,
SGP_READ);
if (error) {
if (error == -EINVAL)
@@ -3477,8 +3480,8 @@ static long shmem_fallocate(struct file
else if (shmem_falloc.nr_unswapped > shmem_falloc.nr_falloced)
error = -ENOMEM;
else
- error = shmem_get_folio(inode, index, &folio,
- SGP_FALLOC);
+ error = shmem_get_folio(inode, index, offset + len,
+ &folio, SGP_FALLOC);
if (error) {
info->fallocend = undo_fallocend;
/* Remove the !uptodate folios we added */
@@ -3829,7 +3832,7 @@ static int shmem_symlink(struct mnt_idma
} else {
inode_nohighmem(inode);
inode->i_mapping->a_ops = &shmem_aops;
- error = shmem_get_folio(inode, 0, &folio, SGP_WRITE);
+ error = shmem_get_folio(inode, 0, 0, &folio, SGP_WRITE);
if (error)
goto out_remove_offset;
inode->i_op = &shmem_symlink_inode_operations;
@@ -3875,7 +3878,7 @@ static const char *shmem_get_link(struct
return ERR_PTR(-ECHILD);
}
} else {
- error = shmem_get_folio(inode, 0, &folio, SGP_READ);
+ error = shmem_get_folio(inode, 0, 0, &folio, SGP_READ);
if (error)
return ERR_PTR(error);
if (!folio)
@@ -5343,7 +5346,7 @@ struct folio *shmem_read_folio_gfp(struc
struct folio *folio;
int error;
- error = shmem_get_folio_gfp(inode, index, &folio, SGP_CACHE,
+ error = shmem_get_folio_gfp(inode, index, 0, &folio, SGP_CACHE,
gfp, NULL, NULL);
if (error)
return ERR_PTR(error);
--- a/mm/userfaultfd.c~mmtmpfs-consider-end-of-file-write-in-shmem_is_huge
+++ a/mm/userfaultfd.c
@@ -391,7 +391,7 @@ static int mfill_atomic_pte_continue(pmd
struct page *page;
int ret;
- ret = shmem_get_folio(inode, pgoff, &folio, SGP_NOALLOC);
+ ret = shmem_get_folio(inode, pgoff, 0, &folio, SGP_NOALLOC);
/* Our caller expects us to return -EFAULT if we failed to find folio */
if (ret == -ENOENT)
ret = -EFAULT;
_
Patches currently in -mm which might be from riel@xxxxxxxxxxx are
mmtmpfs-consider-end-of-file-write-in-shmem_is_huge.patch
