Re: [PATCH]: Clean up of __alloc_pages

[Rohit Seth]

On Sat, 2005-10-29 at 18:47 -0700, Paul Jackson wrote:
> A couple more items:
>  1) Lets try for a consistent use of type "gfp_t" for gfp_mask.
>  2) The can_try_harder flag values were driving me nuts.

Not sure why?  Let me see if some new values could better articulate the
meaning.  Currently if value is < 0 then don't check the watermarks.
When we do check for watermarks, then the value of 1 indicates that it
could go below minimum value.

>  3) The "inline" you added to buffered_rmqueue() blew up my compile.

I will remove the inline based on your and Nick emails.  Though my patch
had inline before the struct....

-static struct page *
-buffered_rmqueue(struct zone *zone, int order, gfp_t gfp_flags)
+static inline struct page *
+buffered_rmqueue(struct zone *zone, int order, gfp_t gfp_flags, int
replenish)

...so that shouldn't have caused any problem.

>  4) The return from try_to_free_pages() was put in "i" for no evident reason.

Will be fixed.

>  5) I have no clue what the replenish flag you added to buffered_rmqueue does.
> 

A bit futuristic.  Will need it when pcp allocations gets checked first
(as Nick also mentioned).  Will remove it for now.

> You patch has:
> > can_try_harder can have following 
> >  * values:
> >  * -1 => No need to check for the watermarks.
> >  *  0 => Don't go too low down in deeps below the low watermark (GFP_HIGH)
> >  *  1 => Go far below the low watermark.  See zone_watermark_ok (RT TASK)
> 
> Later on, you have an inequality test on this value:
> 	if ((can_try_harder >= 0)
> and a non-zero test:
> 	if (can_try_harder)

The last line is from zone_watermark_ok.  The first check is in
get_page_from_freelist.  There is no (can_try_harder) check for this
flag in that function.

> 
> That's three magic values, not even in increasing order of "how hard
> one should try", tested a couple of different ways that requires
> absorbing the complete details of the three values and their ordering
> before one can read the code.
> 



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/