On 12/08/22 at 08:52pm, Uladzislau Rezki wrote:
> On Wed, Dec 07, 2022 at 04:03:41PM +0800, Baoquan He wrote:
......
> > > > @@ -1967,6 +1972,9 @@ static void *new_vmap_block(unsigned int order, gfp_t gfp_mask)
> > > > kfree(vb);
> > > > return ERR_CAST(va);
> > > > }
> > > > + spin_lock(&vmap_area_lock);
> > > > + va->flags = VMAP_RAM|VMAP_BLOCK;
> > > > + spin_unlock(&vmap_area_lock);
> > > >
> > > The per-cpu code was created as a fast per-cpu allocator because of high
> > > vmalloc lock contention. If possible we should avoid of locking of the
> > > vmap_area_lock. Because it has a high contention.
> >
> > Fair enough. I made below draft patch to address the concern. By
> > adding argument va_flags to alloc_vmap_area(), we can pass the
> > vm_map_ram flags into alloc_vmap_area and filled into vmap_area->flags.
> > With this, we don't need add extra action to acquire vmap_area_root lock
> > and do the flags setting. Is it OK to you?
> >
> > From 115f6080b339d0cf9dd20c5f6c0d3121f6b22274 Mon Sep 17 00:00:00 2001
> > From: Baoquan He <bhe@xxxxxxxxxx>
> > Date: Wed, 7 Dec 2022 11:08:14 +0800
> > Subject: [PATCH] mm/vmalloc: change alloc_vmap_area() to pass in va_flags
> >
> > With this change, we can pass and set vmap_area->flags for vm_map_ram area
> > in alloc_vmap_area(). Then no extra action need be added to acquire
> > vmap_area_lock when doing the vmap_area->flags setting.
> >
> > Signed-off-by: Baoquan He <bhe@xxxxxxxxxx>
> > ---
> > mm/vmalloc.c | 13 +++++++++----
> > 1 file changed, 9 insertions(+), 4 deletions(-)
> >
> > diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> > index ccaa461998f3..d74eddec352f 100644
> > --- a/mm/vmalloc.c
> > +++ b/mm/vmalloc.c
> > @@ -1586,7 +1586,9 @@ preload_this_cpu_lock(spinlock_t *lock, gfp_t gfp_mask, int node)
> > static struct vmap_area *alloc_vmap_area(unsigned long size,
> > unsigned long align,
> > unsigned long vstart, unsigned long vend,
> > - int node, gfp_t gfp_mask)
> > + int node, gfp_t gfp_mask,
> > + unsigned long va_flags)
> > +)
> > {
> > struct vmap_area *va;
> > unsigned long freed;
> > @@ -1630,6 +1632,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size,
> > va->va_start = addr;
> > va->va_end = addr + size;
> > va->vm = NULL;
> > + va->flags = va_flags;
> >
> > spin_lock(&vmap_area_lock);
> > insert_vmap_area(va, &vmap_area_root, &vmap_area_list);
> > @@ -1961,7 +1964,8 @@ static void *new_vmap_block(unsigned int order, gfp_t gfp_mask)
> >
> > va = alloc_vmap_area(VMAP_BLOCK_SIZE, VMAP_BLOCK_SIZE,
> > VMALLOC_START, VMALLOC_END,
> > - node, gfp_mask);
> > + node, gfp_mask,
> > + VMAP_RAM|VMAP_BLOCK);
> > if (IS_ERR(va)) {
> > kfree(vb);
> > return ERR_CAST(va);
> > @@ -2258,7 +2262,8 @@ void *vm_map_ram(struct page **pages, unsigned int count, int node)
> > } else {
> > struct vmap_area *va;
> > va = alloc_vmap_area(size, PAGE_SIZE,
> > - VMALLOC_START, VMALLOC_END, node, GFP_KERNEL);
> > + VMALLOC_START, VMALLOC_END,
> > + node, GFP_KERNEL, VMAP_RAM|VMAP_BLOCK);
> > if (IS_ERR(va))
> > return NULL;
> >
> > @@ -2498,7 +2503,7 @@ static struct vm_struct *__get_vm_area_node(unsigned long size,
> > if (!(flags & VM_NO_GUARD))
> > size += PAGE_SIZE;
> >
> > - va = alloc_vmap_area(size, align, start, end, node, gfp_mask);
> > + va = alloc_vmap_area(size, align, start, end, node, gfp_mask, 0);
> > if (IS_ERR(va)) {
> > kfree(area);
> > return NULL;
> > --
> > 2.34.1
> >
> Yes, this is better than it was before. Adding an extra parameter makes
> it more valid and logical.
That's great. I will add this in v2.
