[pve-devel] [RFC PATCH http-server] fix #6230: increase allowed post size

[Dominik Csapak]

On 4/2/25 22:09, Thomas Lamprecht wrote:
> Am 12.03.25 um 14:27 schrieb Dominik Csapak:
>> In some situations, e.g. having a large resource mapping, the UI can
>> generate a request that is bigger than the current limit of 64KiB.
>>
>> Our files in pmxcfs can grow up to 1 MiB, so theoretically, a single
>> mapping can grow to that size. In practice, a single entry will have
>> much less. In #6230, a user has a mapping with about ~130KiB.
>>
>> Increase the limit to 512KiB so we have a bit of buffer left.
> 
> s/buffer/headroom/ ?
> 

yes, makes more sense^^

>>
>> We have to also increase the 'rbuf_max' size here, otherwise the request
>> will fail (since the buffer is too small for the request).
>> Since the post limit and the rbuf_max are tightly coupled, let it
>> reflect that in the code. To do that sum the post size + max header
>> size there.
>>
>> Signed-off-by: Dominik Csapak <d.csapak at proxmox.com>
>> ---
>> sending as RFC because:
>> * not sure about the rbuf_max calculation, but we have to increase it
>>    when we increase $limit_max_post. (not sure how much is needed exactly)
>> * ther are alternative ways to deal with that, but some of those are vastly
>>    more work:
>>    - optimize the pci mapping to reduce the number of bytes we have to
>>      send (e.g. by reducing the property names, or somehow magically
>>      detect devices that belong together)
>>    - add a new api for the mappings that can update the entries without
>>      sending the whole mapping again (not sure if we can make this
>>      backwards compatible)
>>    - ignore the problem and simply tell the users to edit the file
>>      manually (I don't like this one...)
>>
>> also, I tried to benchmark this, but did not find a tool that does this
>> in a good way (e.g. apachebench complained about ssl, and i couldn't get
>> it to work right). @Thomas you did such benchmarks laft according to git
>> log, do you remember what you used then?
> 
> argh, my commit message back then looks like I tried to write what I used
> but then fubmled (or got knocked on the head) and sent it out unfinished.
> To my defence, Wolfgang applied it ;P
> 
> I'm not totally sure what I used back then, might have been something
> custom-made too. FWIW, recently I used oha [0] and found it quite OK, albeit
> I did not try it with POST data, but one can define the method and pass a
> request body from CLI argument directly or a file, and it has a flag to
> allow "insecure" TLS certs.
> 
> [0]: https://github.com/hatoo/oha

thanks, i'll try to do some benchmarks with it

> 
>> @@ -1891,7 +1891,7 @@ sub accept_connections {
>>   	    $self->{conn_count}++;
>>   	    $reqstate->{hdl} = AnyEvent::Handle->new(
>>   		fh => $clientfh,
>> -		rbuf_max => 64*1024,
>> +		rbuf_max => $limit_max_post + ($limit_max_headers * $limit_max_header_size),
> 
> The header part is wrong as the header limits are independent, i.e., the
> request must have less than $limit_max_headers separate headers and all
> those together must be smaller than $limit_max_header_size.
> 
> So just adding $limit_max_header_size is enough, no multiplication required.
> 

ah yes, seems i read those wrong

>>   		timeout => $self->{timeout},
>>   		linger => 0, # avoid problems with ssh - really needed ?
>>   		on_eof   => sub {