On Tue, Sep 26, 2023 at 09:05:49PM +0900, Jaeseon Sim wrote:
> > > > We do not have above code anymore:
> > > Sorry, I tried to say it in a simplified way and it caused a misunderstanding.
> > >
> > > <snip>
> > > static __always_inline int
> > > adjust_va_to_fit_type(struct rb_root *root, struct list_head *head,
> > > struct vmap_area *va, unsigned long nva_start_addr,
> > > unsigned long size)
> > >
> > > } else if (type == NE_FIT_TYPE) {
> > > /*
> > > * Split no edge of fit VA.
> > > *
> > > * | |
> > > * L V NVA V R
> > > * |---|-------|---|
> > > */
> > > lva = __this_cpu_xchg(ne_fit_preload_node, NULL);
> > > if (unlikely(!lva)) {
> > > /*
> > > * For percpu allocator we do not do any pre-allocation
> > > * and leave it as it is. The reason is it most likely
> > > * never ends up with NE_FIT_TYPE splitting. In case of
> > > * percpu allocations offsets and sizes are aligned to
> > > * fixed align request, i.e. RE_FIT_TYPE and FL_FIT_TYPE
> > > * are its main fitting cases.
> > > *
> > > * There are a few exceptions though, as an example it is
> > > * a first allocation (early boot up) when we have "one"
> > > * big free space that has to be split.
> > > *
> > > * Also we can hit this path in case of regular "vmap"
> > > * allocations, if "this" current CPU was not preloaded.
> > > * See the comment in alloc_vmap_area() why. If so, then
> > > * GFP_NOWAIT is used instead to get an extra object for
> > > * split purpose. That is rare and most time does not
> > > * occur.
> > > *
> > > * What happens if an allocation gets failed. Basically,
> > > * an "overflow" path is triggered to purge lazily freed
> > > * areas to free some memory, then, the "retry" path is
> > > * triggered to repeat one more time. See more details
> > > * in alloc_vmap_area() function.
> > > */
> > > lva = kmem_cache_alloc(vmap_area_cachep, GFP_NOWAIT);
> > > if (!lva)
> > > return -1;
> > > }
> > > <snip>
> > >
> > > Above allocation fail will meet WARN_ON_ONCE in the current kernel now.
> > > Should It be handled by alloc_vmap_area()?, as you described in a comment.
> > >
> > WARN_ONCE_ONCE() is a warning and not a panic, though your kernel config
> > considers it as a panic. Right, we go on retry path and we can remove
>
> Right, only in case panic_on_warn is enabled..
>
> > the warning only for GFP_NOWAIT-alloc-error. From the other hand we
> > should still have possibility to trigger a warning if an allocation
> > is not successful: no vmap space or low memory condition, thus no
> > physical memory.
>
> Yes, but GFP_NOWAIT-alloc-error can easily occur for low-memory device.
>
Agree. You are really in a low memory condition. We end up here only if
pre-loading also has not succeeded, i.e. GFP_KERNEL also fails.
But i agree with you, we should "improve the warning" because we drain
and repeat.
> How about changing fix as below?:
>
> <snip>
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -1468,6 +1468,7 @@ adjust_va_to_fit_type(struct rb_root *root, struct list_head *head,
> */
> va->va_start = nva_start_addr + size;
> } else {
> + WARN_ON_ONCE(1);
> return -1;
> }
>
> @@ -1522,7 +1523,7 @@ __alloc_vmap_area(struct rb_root *root, struct list_head *head,
>
> /* Update the free vmap_area. */
> ret = adjust_va_to_fit_type(root, head, va, nva_start_addr, size);
> - if (WARN_ON_ONCE(ret))
> + if (ret)
> return vend;
>
> #if DEBUG_AUGMENT_LOWEST_MATCH_CHECK
> @@ -4143,7 +4144,7 @@ struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
> ret = adjust_va_to_fit_type(&free_vmap_area_root,
> &free_vmap_area_list,
> va, start, size);
> - if (WARN_ON_ONCE(unlikely(ret)))
> + if (unlikely(ret))
> /* It is a BUG(), but trigger recovery instead. */
> goto recovery;
>
> <snip>
> It will WARN_ONCE_ONCE() only if classify_va_fit_type() is "(type == NOTHING_FIT)".
>
This is good but i think it should be improved further. We need to
understand from the warning when no memory and when there is no a
vmap space, so:
- if NOTHING_FIT, we should WARN() for sure;
- Second place in the pcpu_get_vm_area(), we do not use NE_FIT. Only in
the begging after boot, but potentially we can trigger -ENOMEM and we
should warn in this case. Otherwise you just hide it;
- And last one if after repeating we still do not manage to allocate.
--
Uladzislau Rezki
