Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] xen/balloon: Support xend-based toolstack take two
To:     Jan Beulich <jbeulich@suse.com>
Cc:     xen-devel@lists.xenproject.org, linux-kernel@vger.kernel.org,
        Stefano Stabellini <sstabellini@kernel.org>,
        Boris Ostrovsky <boris.ostrovsky@oracle.com>,
        stable@vger.kernel.org
References: <20200116170004.14373-1-jgross@suse.com>
 <c29c92e3-eb20-7e0a-0174-ef72398b0998@suse.com>
From:   =?UTF-8?B?SsO8cmdlbiBHcm/Dnw==?= <jgross@suse.com>
Message-ID: <dc509037-a7d6-caa5-8000-28aeb20b638e@suse.com>
Date:   Fri, 17 Jan 2020 12:31:40 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.2.1
MIME-Version: 1.0
In-Reply-To: <c29c92e3-eb20-7e0a-0174-ef72398b0998@suse.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 17.01.20 12:01, Jan Beulich wrote:
> On 16.01.2020 18:00, Juergen Gross wrote:
>> Commit 3aa6c19d2f38be ("xen/balloon: Support xend-based toolstack")
>> tried to fix a regression with running on rather ancient Xen versions.
>> Unfortunately the fix was based on the assumption that xend would
>> just use another Xenstore node, but in reality only some downstream
>> versions of xend are doing that. The upstream xend does not write
>> that Xenstore node at all, so the problem must be fixed in another
>> way.
>>
>> The easiest way to achieve that is to fall back to the behavior before
>> commit 5266b8e4445c ("xen: fix booting ballooned down hvm guest")
>> in case the static memory maximum can't be read.
> 
> I could use some help here: Prior to said commit there was
> 
> 	target_diff = new_target - balloon_stats.target_pages;
> 
> 
> Which is, afaict, ...
> 
>> --- a/drivers/xen/xen-balloon.c
>> +++ b/drivers/xen/xen-balloon.c
>> @@ -94,7 +94,7 @@ static void watch_target(struct xenbus_watch *watch,
>>   				  "%llu", &static_max) == 1))
>>   			static_max >>= PAGE_SHIFT - 10;
>>   		else
>> -			static_max = new_target;
>> +			static_max = balloon_stats.current_pages;
>>   
>>   		target_diff = (xen_pv_domain() || xen_initial_domain()) ? 0
>>   				: static_max - balloon_stats.target_pages;
> 
> ... what the code does before your change. Afaict there was
> never a use of balloon_stats.current_pages in this function.

That is a little bit indirect, yes. In the end I want static_max to
be either the maximum reported by Xen, or if not available, the current
assumed memory size, which can be found in balloon_stats.current_pages.

The main idea is to avoid a negative target_diff which would result in
not ballooning down.


Juergen