Re: [PATCH 2/7] mm: introduce kvmalloc and kvmalloc_node

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




On Wed, 8 Jul 2015, Andrew Morton wrote:

> On Wed, 8 Jul 2015 19:03:08 -0400 (EDT) Mikulas Patocka <mpatocka@xxxxxxxxxx> wrote:
> 
> > 
> > 
> > On Tue, 7 Jul 2015, Andrew Morton wrote:
> > 
> > > On Tue, 7 Jul 2015 11:10:09 -0400 (EDT) Mikulas Patocka <mpatocka@xxxxxxxxxx> wrote:
> > > 
> > > > Introduce the functions kvmalloc and kvmalloc_node. These functions
> > > > provide reliable allocation of object of arbitrary size. They attempt to
> > > > do allocation with kmalloc and if it fails, use vmalloc. Memory allocated
> > > > with these functions should be freed with kvfree.
> > > 
> > > Sigh.  We've resisted doing this because vmalloc() is somewhat of a bad
> > > thing, and we don't want to make it easy for people to do bad things.
> > > 
> > > And vmalloc is bad because a) it's slow and b) it does GFP_KERNEL
> > > allocations for page tables and c) it is susceptible to arena
> > > fragmentation.
> > 
> > This patch makes less use of vmalloc.
> > 
> > The typical pattern is that someone notices random failures due to memory 
> > fragmentation in some subsystem that uses large kmalloc - so he replaces 
> > kmalloc with vmalloc - and the code gets slower because of that. With this 
> > patch, you can replace many vmalloc users with kvmalloc - and vmalloc will 
> > be used only very rarely, when the memory is too fragmented for kmalloc.
> 
> Yes, I guess there is that.
> 
> > Here I'm sending next version of the patch with comments added.
> 
> You didn't like kvzalloc()?  We can always add those later...
> 
> > --- linux-4.2-rc1.orig/include/linux/mm.h	2015-07-07 15:58:11.000000000 +0200
> > +++ linux-4.2-rc1/include/linux/mm.h	2015-07-08 19:22:24.000000000 +0200
> > @@ -400,6 +400,11 @@ static inline int is_vmalloc_or_module_a
> >  }
> >  #endif
> >  
> > +extern void *kvmalloc_node(size_t size, gfp_t gfp, int node);
> > +static inline void *kvmalloc(size_t size, gfp_t gfp)
> > +{
> > +	return kvmalloc_node(size, gfp, NUMA_NO_NODE);
> > +}
> >  extern void kvfree(const void *addr);
> >  
> >  static inline void compound_lock(struct page *page)
> > Index: linux-4.2-rc1/mm/util.c
> > ===================================================================
> > --- linux-4.2-rc1.orig/mm/util.c	2015-07-07 15:58:11.000000000 +0200
> > +++ linux-4.2-rc1/mm/util.c	2015-07-08 19:22:26.000000000 +0200
> > @@ -316,6 +316,61 @@ unsigned long vm_mmap(struct file *file,
> >  }
> >  EXPORT_SYMBOL(vm_mmap);
> >  
> > +void *kvmalloc_node(size_t size, gfp_t gfp, int node)
> > +{
> > +	void *p;
> > +	unsigned uninitialized_var(noio_flag);
> > +
> > +	/* vmalloc doesn't support no-wait allocations */
> > +	WARN_ON_ONCE(!(gfp & __GFP_WAIT));
> > +
> > +	if (likely(size <= KMALLOC_MAX_SIZE)) {
> > +		/*
> > +		 * Use __GFP_NORETRY so that we don't loop waiting for the
> > +		 *	allocation - we don't have to loop here, if the memory
> > +		 *	is too fragmented, we fallback to vmalloc.
> 
> I'm not sure about this decision.  The direct reclaim retry code is the
> normal default behaviour and becomes more important with larger allocation
> attempts.  So why turn it off, and make it more likely that we return
> vmalloc memory?

It can avoid triggering the OOM killer in case of fragmented memory.

This is general question - if the code can handle allocation failure 
gracefully, what gfp flags should it use? Maybe add some flag 
__GFP_MAYFAIL instead of __GFP_NORETRY that changes the behavior in 
desired way?

Mikulas

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>



[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]