[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20180506185532.GA13604@bombadil.infradead.org>
Date: Sun, 6 May 2018 11:55:32 -0700
From: Matthew Wilcox <willy@...radead.org>
To: Huaisheng HS1 Ye <yehs1@...ovo.com>
Cc: Michal Hocko <mhocko@...nel.org>,
"akpm@...ux-foundation.org" <akpm@...ux-foundation.org>,
"linux-mm@...ck.org" <linux-mm@...ck.org>,
"vbabka@...e.cz" <vbabka@...e.cz>,
"mgorman@...hsingularity.net" <mgorman@...hsingularity.net>,
"pasha.tatashin@...cle.com" <pasha.tatashin@...cle.com>,
"alexander.levin@...izon.com" <alexander.levin@...izon.com>,
"hannes@...xchg.org" <hannes@...xchg.org>,
"penguin-kernel@...ove.SAKURA.ne.jp"
<penguin-kernel@...ove.SAKURA.ne.jp>,
"colyli@...e.de" <colyli@...e.de>,
NingTing Cheng <chengnt@...ovo.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [External] Re: [PATCH 2/3] include/linux/gfp.h: use unsigned
int in gfp_zone
On Sun, May 06, 2018 at 04:17:06PM +0000, Huaisheng HS1 Ye wrote:
> Upload my current patch and testing platform info for reference. This patch has been tested
> on a two sockets platform.
Thank you!
> It works, but some drivers or subsystem shall be modified to fit
> these new type __GFP flags.
> They use these flags directly to realize bit manipulations like this
> below.
>
> eg.
> swiotlb-xen.c (drivers\xen): flags &= ~(__GFP_DMA | __GFP_HIGHMEM);
> extent_io.c (fs\btrfs): mask &= ~(__GFP_DMA32|__GFP_HIGHMEM);
>
> Because of these flags have been encoded within this patch, the
> above operations can cause problem.
I don't think this actually causes problems. At least, no additional
problems. These users will successfully clear __GFP_DMA and __GFP_HIGHMEM
no matter what values GFP_DMA and GFP_HIGHMEM have; the only problem will
be if someone calls them with a zone type they're not expecting (eg DMA32
for the first one or DMA for the second; or MOVABLE for either of them).
The thing is, they're already buggy in those circumstances.
> */
> -#define __GFP_DMA ((__force gfp_t)___GFP_DMA)
> -#define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
> -#define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
> +#define __GFP_DMA ((__force gfp_t)OPT_ZONE_DMA ^ ZONE_NORMAL)
> +#define __GFP_HIGHMEM ((__force gfp_t)ZONE_MOVABLE ^ ZONE_NORMAL)
> +#define __GFP_DMA32 ((__force gfp_t)OPT_ZONE_DMA32 ^ ZONE_NORMAL)
> #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */
[...]
> static inline enum zone_type gfp_zone(gfp_t flags)
> {
> enum zone_type z;
> - int bit = (__force int) (flags & GFP_ZONEMASK);
> + z = ((__force unsigned int)flags & ___GFP_ZONE_MASK) ^ ZONE_NORMAL;
>
> - z = (GFP_ZONE_TABLE >> (bit * GFP_ZONES_SHIFT)) &
> - ((1 << GFP_ZONES_SHIFT) - 1);
> - VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
> + if (z > OPT_ZONE_HIGHMEM) {
> + z = OPT_ZONE_HIGHMEM +
> + !!((__force unsigned int)flags & ___GFP_MOVABLE);
> + }
> return z;
> }
How about:
+#define __GFP_HIGHMEM ((__force gfp_t)OPT_ZONE_HIGHMEM ^ ZONE_NORMAL)
-#define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */
+#define __GFP_MOVABLE ((__force gfp_t)ZONE_MOVABLE ^ ZONE_NORMAL | \
+ ___GFP_MOVABLE)
Then I think you can just make it:
static inline enum zone_type gfp_zone(gfp_t flags)
{
return ((__force int)flags & ___GFP_ZONE_MASK) ^ ZONE_NORMAL;
}
> @@ -370,42 +368,15 @@ static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
> #error GFP_ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
> #endif
You should be able to delete GFP_ZONES_SHIFT too.
Powered by blists - more mailing lists