mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Bannier <bbann...@apache.org>
Subject Re: Review Request 71947: Handled embedded null bytes in abstract domain socket names.
Date Mon, 06 Jan 2020 12:59:19 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/71947/#review219127
-----------------------------------------------------------




3rdparty/libprocess/include/process/address.hpp
Lines 224 (patched)
<https://reviews.apache.org/r/71947/#comment307224>

    Let's include `<cstddef>` for `offsetof`.



3rdparty/libprocess/include/process/address.hpp
Line 226 (original), 227-228 (patched)
<https://reviews.apache.org/r/71947/#comment307226>

    This is not very clear to me. Can we give a very explicit example on how to obtain `_length`,
or alternatively compute it ourself as suggested below?



3rdparty/libprocess/include/process/address.hpp
Lines 238 (patched)
<https://reviews.apache.org/r/71947/#comment307229>

    As discussed offline, let's add a `CHECK` here that `sun_path` is not empty.



3rdparty/libprocess/include/process/address.hpp
Line 232 (original), 247 (patched)
<https://reviews.apache.org/r/71947/#comment307227>

    For unnamed sockets we would now return an empty string here which seems leaky. What do
you think about changing the return type of `Address::path` to e.g., `Option<string>`
instead and returning a `None` in that case?



3rdparty/libprocess/include/process/address.hpp
Lines 275 (patched)
<https://reviews.apache.org/r/71947/#comment307228>

    nit: Missing trailing dot.



3rdparty/libprocess/include/process/address.hpp
Lines 318 (patched)
<https://reviews.apache.org/r/71947/#comment307230>

    Does it make sense to pass an `Option<socklen_t>` defaulted to `None` to make it
clearer that `length` is only interesting in certain cases?
    
    In any case we need to do add documentation on the semantics of the parameters.


- Benjamin Bannier


On Jan. 3, 2020, 2:38 a.m., Benno Evers wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/71947/
> -----------------------------------------------------------
> 
> (Updated Jan. 3, 2020, 2:38 a.m.)
> 
> 
> Review request for mesos, Benjamin Bannier and Benjamin Mahler.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Address handling code for unix domain sockets assumed that
> strlen() could be used to compute the name of a unix domain
> socket, but that fails for unnamed sockets or in the case
> where an abstract domain socket contains embedded null bytes.
> 
> This patch adds a new `length` parameter to correctly handle
> these special cases.
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/include/process/address.hpp 749498056b52b916dfaf6c85f83ecc05e0d5406c

>   3rdparty/libprocess/include/process/network.hpp 8f48a4a78557309a9b1b00d7defb45eed454b077

> 
> 
> Diff: https://reviews.apache.org/r/71947/diff/1/
> 
> 
> Testing
> -------
> 
> Ran existing unit tests and verified that the newly added `CHECK()` doesn't trigger.
> 
> 
> Thanks,
> 
> Benno Evers
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message