From: Markus Armbruster <armbru@redhat.com>
References: <1371106939-6968-1-git-send-email-armbru@redhat.com>
	<1371106939-6968-8-git-send-email-armbru@redhat.com>
	<51B9ED17.4060305@twiddle.net>
Date: Thu, 13 Jun 2013 19:27:13 +0200
In-Reply-To: <51B9ED17.4060305@twiddle.net> (Richard Henderson's message of
	"Thu, 13 Jun 2013 09:02:31 -0700")
Message-ID: <87obb9ly4u.fsf@blackfin.pond.sub.org>
MIME-Version: 1.0
Content-Type: text/plain
Subject: Re: [Qemu-devel] [PATCH RFC 7/8] exec: Don't abort when we can't
	allocate guest memory
To: Richard Henderson <rth@twiddle.net>
Cc: stefano.stabellini@eu.citrix.com, mtosatti@redhat.com, agraf@suse.de, qemu-devel@nongnu.org, borntraeger@de.ibm.com, pbonzini@redhat.com

Richard Henderson <rth@twiddle.net> writes:

> On 06/13/2013 12:02 AM, Markus Armbruster wrote:
>> @@ -945,7 +952,7 @@ static void *file_ram_alloc(RAMBlock *block,
>>      area = mmap(0, memory, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
>>  #endif
>>      if (area == MAP_FAILED) {
>> -        perror("file_ram_alloc: can't mmap RAM pages");
>> +        no_guest_mem(block);
>>          close(fd);
>>          return (NULL);
>>      }
>
> Dead code after no_guest_mem.

Right you are.

Before: if allocation obeying -mem-path fails, we retry without
-mem-path.  Some failures are silent.

After: if mmap() fails, we give up.  Functional change not mentioned in
commit message.  I'll either revert it or document it.

I'm not sure falling back to non-mem-path allocation is a good idea in
all failure cases.

Thanks!