Re: [PATCH v10 4/4] mm/madvise: optimize lazyfreeing with mTHP in madvise_free

From: Barry Song
Date: Thu May 23 2024 - 18:20:32 EST

Next message: Khazhismel Kumykov: "[PATCH v2 1/2] iscsi_tcp: do not bind sockets that already have extra callbacks"
Previous message: Peter Xu: "[PATCH rfcv1.1] mm/x86/pat: Do proper PAT bit shift for large mappings"
Next in thread: Lance Yang: "Re: [PATCH v10 4/4] mm/madvise: optimize lazyfreeing with mTHP in madvise_free"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Fri, Apr 19, 2024 at 1:44 AM Lance Yang <ioworker0@xxxxxxxxx> wrote:
>
> This patch optimizes lazyfreeing with PTE-mapped mTHP[1]
> (Inspired by David Hildenbrand[2]). We aim to avoid unnecessary folio
> splitting if the large folio is fully mapped within the target range.
>
> If a large folio is locked or shared, or if we fail to split it, we just
> leave it in place and advance to the next PTE in the range. But note that
> the behavior is changed; previously, any failure of this sort would cause
> the entire operation to give up. As large folios become more common,
> sticking to the old way could result in wasted opportunities.
>
> On an Intel I5 CPU, lazyfreeing a 1GiB VMA backed by PTE-mapped folios of
> the same size results in the following runtimes for madvise(MADV_FREE) in
> seconds (shorter is better):
>
> Folio Size | Old | New | Change
> ------------------------------------------
> 4KiB | 0.590251 | 0.590259 | 0%
> 16KiB | 2.990447 | 0.185655 | -94%
> 32KiB | 2.547831 | 0.104870 | -95%
> 64KiB | 2.457796 | 0.052812 | -97%
> 128KiB | 2.281034 | 0.032777 | -99%
> 256KiB | 2.230387 | 0.017496 | -99%
> 512KiB | 2.189106 | 0.010781 | -99%
> 1024KiB | 2.183949 | 0.007753 | -99%
> 2048KiB | 0.002799 | 0.002804 | 0%
>
> [1] https://lkml.kernel.org/r/20231207161211.2374093-5-ryan.roberts@xxxxxxx
> [2] https://lore.kernel.org/linux-mm/20240214204435.167852-1-david@redhatcom
>
> Reviewed-by: Ryan Roberts <ryan.roberts@xxxxxxx>
> Acked-by: David Hildenbrand <david@xxxxxxxxxx>
> Signed-off-by: Lance Yang <ioworker0@xxxxxxxxx>
> ---

Hi Lance,
I am getting kernel panic with this patch,

/ # /home/barry/develop/linux/madvfree
[ 78.345305] watchdog: BUG: soft lockup - CPU#3 stuck for 22s! [madvfree:101]
[ 78.345992] Modules linked in:
[ 78.346942] irq event stamp: 0
[ 78.347311] hardirqs last enabled at (0): [<0000000000000000>] 0x0
[ 78.348407] hardirqs last disabled at (0): [<ffff8000800add04>]
copy_process+0x654/0x19a8
[ 78.349291] softirqs last enabled at (0): [<ffff8000800add04>]
copy_process+0x654/0x19a8
[ 78.349851] softirqs last disabled at (0): [<0000000000000000>] 0x0
[ 78.350544] CPU: 3 PID: 101 Comm: madvfree Not tainted
6.9.0-ge51ae633c861 #253
[ 78.351200] Hardware name: linux,dummy-virt (DT)
[ 78.351747] pstate: 01400005 (nzcv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
[ 78.352314] pc : queued_spin_lock_slowpath+0x5c/0x528
[ 78.352772] lr : do_raw_spin_lock+0xc8/0x120
[ 78.353245] sp : ffff8000863d3720
[ 78.353657] x29: ffff8000863d3720 x28: ffff0000c45a8ff8 x27: 0800000103b24003
[ 78.354632] x26: ffff0000c3b26080 x25: fffffdffc0000000 x24: ffff8000822c2d10
[ 78.355446] x23: ffff80008403018f x22: ffff8000863d38e0 x21: 0000ffff7f000000
[ 78.356259] x20: ffff800082fbe008 x19: ffff0000c3b26080 x18: 0000000000000000
[ 78.357120] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
[ 78.357967] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000
[ 78.358878] x11: ff7ffffffffffbff x10: 0040000000000041 x9 : ffff800080143750
[ 78.359818] x8 : ffff8000863d3708 x7 : 0000000000000000 x6 : ffff8000803b34e8
[ 78.360688] x5 : 0000000000000000 x4 : 0000000000000001 x3 : ffff800082fbe008
[ 78.361602] x2 : ffff80012ac1f000 x1 : 0000000000000000 x0 : 0000000000000080
[ 78.362755] Call trace:
[ 78.363132] queued_spin_lock_slowpath+0x5c/0x528
[ 78.363614] do_raw_spin_lock+0xc8/0x120
[ 78.364041] _raw_spin_lock+0x58/0x70
[ 78.364455] __pte_offset_map_lock+0x98/0x210
[ 78.364891] madvise_free_pte_range+0x88/0xa58
[ 78.365406] walk_pgd_range+0x390/0x808
[ 78.365829] __walk_page_range+0x1e0/0x1f0
[ 78.366256] walk_page_range+0x1f0/0x2c8
[ 78.366676] madvise_free_single_vma+0x16c/0x308
[ 78.367115] madvise_vma_behavior+0x504/0xa20
[ 78.367549] madvise_walk_vmas+0xc0/0x128
[ 78.367972] do_madvise.part.0+0x110/0x558
[ 78.368398] __arm64_sys_madvise+0x68/0x88
[ 78.368826] invoke_syscall+0x50/0x128
[ 78.369332] el0_svc_common.constprop.0+0x48/0xf8
[ 78.369778] do_el0_svc+0x28/0x40
[ 78.370184] el0_svc+0x50/0x150
[ 78.370583] el0t_64_sync_handler+0x13c/0x158
[ 78.371017] el0t_64_sync+0x1a4/0x1a8
[ 102.345217] watchdog: BUG: soft lockup - CPU#3 stuck for 45s! [madvfree:101]
[ 102.345835] Modules linked in:
[ 102.346290] irq event stamp: 0
[ 102.346715] hardirqs last enabled at (0): [<0000000000000000>] 0x0
[ 102.347252] hardirqs last disabled at (0): [<ffff8000800add04>]
copy_process+0x654/0x19a8
[ 102.347796] softirqs last enabled at (0): [<ffff8000800add04>]
copy_process+0x654/0x19a8
[ 102.348333] softirqs last disabled at (0): [<0000000000000000>] 0x0
[ 102.348925] CPU: 3 PID: 101 Comm: madvfree Tainted: G L
6.9.0-ge51ae633c861 #253
[ 102.349549] Hardware name: linux,dummy-virt (DT)
[ 102.349988] pstate: 01400005 (nzcv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
[ 102.350535] pc : queued_spin_lock_slowpath+0x5c/0x528
[ 102.351010] lr : do_raw_spin_lock+0xc8/0x120
[ 102.351508] sp : ffff8000863d3720
[ 102.351939] x29: ffff8000863d3720 x28: ffff0000c45a8ff8 x27: 0800000103b24003
[ 102.352811] x26: ffff0000c3b26080 x25: fffffdffc0000000 x24: ffff8000822c2d10
[ 102.353772] x23: ffff80008403018f x22: ffff8000863d38e0 x21: 0000ffff7f000000
[ 102.354625] x20: ffff800082fbe008 x19: ffff0000c3b26080 x18: 0000000000000000
[ 102.355495] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
[ 102.356370] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000000
[ 102.357333] x11: ff7ffffffffffbff x10: 0040000000000041 x9 : ffff800080143750
[ 102.358273] x8 : ffff8000863d3708 x7 : 0000000000000000 x6 : ffff8000803b34e8
[ 102.359112] x5 : 0000000000000000 x4 : 0000000000000001 x3 : ffff800082fbe008
[ 102.360001] x2 : ffff80012ac1f000 x1 : 0000000000000000 x0 : 0000000000000080
[ 102.360887] Call trace:
[ 102.361289] queued_spin_lock_slowpath+0x5c/0x528
[ 102.361768] do_raw_spin_lock+0xc8/0x120
[ 102.362294] _raw_spin_lock+0x58/0x70
[ 102.362825] __pte_offset_map_lock+0x98/0x210
[ 102.363299] madvise_free_pte_range+0x88/0xa58
[ 102.363771] walk_pgd_range+0x390/0x808
[ 102.364268] __walk_page_range+0x1e0/0x1f0
[ 102.364729] walk_page_range+0x1f0/0x2c8
[ 102.365263] madvise_free_single_vma+0x16c/0x308
[ 102.365786] madvise_vma_behavior+0x504/0xa20
[ 102.366315] madvise_walk_vmas+0xc0/0x128
[ 102.366779] do_madvise.part.0+0x110/0x558
[ 102.367269] __arm64_sys_madvise+0x68/0x88
[ 102.367714] invoke_syscall+0x50/0x128
[ 102.368110] el0_svc_common.constprop.0+0x48/0xf8
[ 102.368574] do_el0_svc+0x28/0x40
[ 102.369001] el0_svc+0x50/0x150
[ 102.369464] el0t_64_sync_handler+0x13c/0x158
[ 102.369916] el0t_64_sync+0x1a4/0x1a8
[ 126.345236] watchdog: BUG: soft lockup - CPU#3 stuck for 67s! [madvfree:101]

the test code is as belows,

#define MADV_PAGEOUT 21 /* reclaim these pages */

#define DATA_SIZE (128UL * 1024 * 1024)
#define PAGE_SIZE (4UL * 1024)
#define LARGE_FOLIO_SIZE (64UL * 1024)

static void *read_data(void *addr)
{
unsigned long i;

for (i = 0; i < DATA_SIZE * 2; i += PAGE_SIZE) {
if (*((char *)addr + i) != (char)i) {
}
}
}

static void *lazyfree(void *addr)
{
unsigned long i;

madvise(addr, DATA_SIZE * 2, MADV_FREE);
}

int main(int argc, char **argv)
{
void *addr = mmap(NULL, DATA_SIZE * 2, PROT_READ | PROT_WRITE,
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
memset(addr, 0x11, DATA_SIZE * 2);
lazyfree(addr);

while(1) {
sleep(1);
read_data(addr);
}
return 0;
}

> mm/madvise.c | 85 +++++++++++++++++++++++++++-------------------------
> 1 file changed, 44 insertions(+), 41 deletions(-)
>
> diff --git a/mm/madvise.c b/mm/madvise.c
> index 4597a3568e7e..ed125ad8a21e 100644
> --- a/mm/madvise.c
> +++ b/mm/madvise.c
> @@ -643,6 +643,7 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
> unsigned long end, struct mm_walk *walk)
>
> {
> + const cydp_t cydp_flags = CYDP_CLEAR_YOUNG | CYDP_CLEAR_DIRTY;
> struct mmu_gather *tlb = walk->private;
> struct mm_struct *mm = tlb->mm;
> struct vm_area_struct *vma = walk->vma;
> @@ -697,44 +698,57 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
> continue;
>
> /*
> - * If pmd isn't transhuge but the folio is large and
> - * is owned by only this process, split it and
> - * deactivate all pages.
> + * If we encounter a large folio, only split it if it is not
> + * fully mapped within the range we are operating on. Otherwise
> + * leave it as is so that it can be marked as lazyfree. If we
> + * fail to split a folio, leave it in place and advance to the
> + * next pte in the range.
> */
> if (folio_test_large(folio)) {
> - int err;
> + bool any_young, any_dirty;
>
> - if (folio_likely_mapped_shared(folio))
> - break;
> - if (!folio_trylock(folio))
> - break;
> - folio_get(folio);
> - arch_leave_lazy_mmu_mode();
> - pte_unmap_unlock(start_pte, ptl);
> - start_pte = NULL;
> - err = split_folio(folio);
> - folio_unlock(folio);
> - folio_put(folio);
> - if (err)
> - break;
> - start_pte = pte =
> - pte_offset_map_lock(mm, pmd, addr, &ptl);
> - if (!start_pte)
> - break;
> - arch_enter_lazy_mmu_mode();
> - pte--;
> - addr -= PAGE_SIZE;
> - continue;
> + nr = madvise_folio_pte_batch(addr, end, folio, pte,
> + ptent, &any_young, &any_dirty);
> +
> + if (nr < folio_nr_pages(folio)) {
> + int err;
> +
> + if (folio_likely_mapped_shared(folio))
> + continue;
> + if (!folio_trylock(folio))
> + continue;
> + folio_get(folio);
> + arch_leave_lazy_mmu_mode();
> + pte_unmap_unlock(start_pte, ptl);
> + start_pte = NULL;
> + err = split_folio(folio);
> + folio_unlock(folio);
> + folio_put(folio);
> + pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
> + start_pte = pte;
> + if (!start_pte)
> + break;
> + arch_enter_lazy_mmu_mode();
> + if (!err)
> + nr = 0;
> + continue;
> + }
> +
> + if (any_young)
> + ptent = pte_mkyoung(ptent);
> + if (any_dirty)
> + ptent = pte_mkdirty(ptent);
> }
>
> if (folio_test_swapcache(folio) || folio_test_dirty(folio)) {
> if (!folio_trylock(folio))
> continue;
> /*
> - * If folio is shared with others, we mustn't clear
> - * the folio's dirty flag.
> + * If we have a large folio at this point, we know it is
> + * fully mapped so if its mapcount is the same as its
> + * number of pages, it must be exclusive.
> */
> - if (folio_mapcount(folio) != 1) {
> + if (folio_mapcount(folio) != folio_nr_pages(folio)) {
> folio_unlock(folio);
> continue;
> }
> @@ -750,19 +764,8 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr,
> }
>
> if (pte_young(ptent) || pte_dirty(ptent)) {
> - /*
> - * Some of architecture(ex, PPC) don't update TLB
> - * with set_pte_at and tlb_remove_tlb_entry so for
> - * the portability, remap the pte with old|clean
> - * after pte clearing.
> - */
> - ptent = ptep_get_and_clear_full(mm, addr, pte,
> - tlb->fullmm);
> -
> - ptent = pte_mkold(ptent);
> - ptent = pte_mkclean(ptent);
> - set_pte_at(mm, addr, pte, ptent);
> - tlb_remove_tlb_entry(tlb, pte, addr);
> + clear_young_dirty_ptes(vma, addr, pte, nr, cydp_flags);
> + tlb_remove_tlb_entries(tlb, pte, nr, addr);
> }
> folio_mark_lazyfree(folio);
> }
> --
> 2.33.1
>

Next message: Khazhismel Kumykov: "[PATCH v2 1/2] iscsi_tcp: do not bind sockets that already have extra callbacks"
Previous message: Peter Xu: "[PATCH rfcv1.1] mm/x86/pat: Do proper PAT bit shift for large mappings"
Next in thread: Lance Yang: "Re: [PATCH v10 4/4] mm/madvise: optimize lazyfreeing with mTHP in madvise_free"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]