[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] xen-mapcache: use MAP_FIXED flag so the mmap ad
From: |
Igor Druzhinin |
Subject: |
Re: [Qemu-devel] [PATCH] xen-mapcache: use MAP_FIXED flag so the mmap address hint is always honored |
Date: |
Fri, 15 Mar 2019 20:24:06 +0000 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.5.1 |
On 15/03/2019 18:38, Anthony PERARD wrote:
> On Fri, Mar 15, 2019 at 06:14:09PM +0000, Anthony PERARD wrote:
>> On Fri, Mar 15, 2019 at 09:58:47AM +0100, Roger Pau Monne wrote:
>>> Or if it's not possible to honor the hinted address an error is returned
>>> instead. This makes it easier to spot the actual failure, instead of
>>> failing later on when the caller of xen_remap_bucket realizes the
>>> mapping has not been created at the requested address.
>>>
>>> Also note that at least on FreeBSD using MAP_FIXED will cause mmap to
>>> try harder to honor the passed address.
>>>
>>> Signed-off-by: Roger Pau Monné <address@hidden>
>>
>> The patch looks fine, and MAP_FIXED seems to be the expected behavior
>> even on Linux.
>>
>> Acked-by: Anthony PERARD <address@hidden>
>>
>>> diff --git a/hw/i386/xen/xen-mapcache.c b/hw/i386/xen/xen-mapcache.c
>>> @@ -185,8 +185,14 @@ static void xen_remap_bucket(MapCacheEntry *entry,
>>> }
>>>
>>> if (!dummy) {
>>> + /*
>>> + * If the caller has requested the mapping at a specific address
>>> use
>>> + * MAP_FIXED to make sure it's honored. Note that with MAP_FIXED at
>>> + * least FreeBSD will try harder to honor the passed address.
>>> + */
>>> vaddr_base = xenforeignmemory_map2(xen_fmem, xen_domid, vaddr,
>>> - PROT_READ | PROT_WRITE, 0,
>>> + PROT_READ | PROT_WRITE,
>>> + vaddr ? MAP_FIXED : 0,
>>> nb_pfn, pfns, err);
>
> I've noticed that there's a mmap call just after which has also vaddr as
> hint, should that second call also have MAP_FIXED as flags?
I think so, yes. The intended usage of xen_remap_bucket() is to create a
mapcache entry - either dummy or real. So this patch fixes the present
problem for real entry now but not for dummy. In future, there might be
xen_remap_bucket() calls to create dummy entries at a predefined
location so, I think, MAP_FIXED should be passed to mmap() call as well.
Igor