Tarantool development patches archive
 help / color / mirror / Atom feed
From: Cyrill Gorcunov <gorcunov@gmail.com>
To: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
Cc: tml <tarantool-patches@dev.tarantool.org>
Subject: Re: [Tarantool-patches] [PATCH 2/2] raft: decode even invalid states of raft
Date: Wed, 28 Oct 2020 01:36:30 +0300	[thread overview]
Message-ID: <20201027223630.GC2093@grain> (raw)
In-Reply-To: <02839650-d213-98d2-1cbf-7fb95d52102c@tarantool.org>

On Tue, Oct 27, 2020 at 11:18:18PM +0100, Vladislav Shpilevoy wrote:
> >  lbox_info_election(struct lua_State *L)
> >  {
> >  	lua_createtable(L, 0, 4);
> > -	lua_pushstring(L, raft_state_strs[raft.state]);
> > +	lua_pushstring(L, raft_state_str(raft.state));
> 
> 1. Raft.state is always valid. It can't be NONE. If a request with a
> bad state is received, it just should be rejected with a protocol
> error.

You miss the main point -- the state string decoder *must* accept
any value instead of relaying on some other code which should filter
bad states. I even put a comment in the code, letme repeat

 enum raft_state {
+	/**
+	 * Unknown state, never valid, for sequence completeness
+	 * and to print out wrong state from incoming packets.
+	 */
+	RAFT_STATE_NONE = 0,

> >  /**
> >   * Check if Raft is completely synced with disk. Meaning all its critical values
> >   * are in WAL. Only in that state the node can become a leader or a candidate.
> > @@ -284,7 +297,7 @@ raft_request_to_string(const struct raft_request *req)
> >  	}
> >  	if (req->state != 0) {
> >  		rc = snprintf(pos, size, ", state: %s",
> > -			      raft_state_strs[req->state]);
> > +			      raft_state_str(req->state));
> 
> 2. By doing so you didn't fix the actual issue - the state is still applied in
> raft_process_msg(). Moreover, in debug build it will crash, because there is
> a switch-case by the received state, which does panic() on an unknown state.

And how it is related to the request to make raft state decode safe? Seriously
I don't get it -- the raft state string was a global variable now it is a function
which implies it simply must be ready for *any* bloody state, even invalid ones.

> 
> The invalid state should be filtered out in the beginning of raft_process_msg(),
> like all the other fields checked there for sanity. State is checked here, but
> only for not being 0. Need to check for the max value too, that is it.

No, either provide a decode function which can take any argument or make it
static and never export to another tarantool code. I took the first approach.

Anyway, this is your code so I won't insist, up to you.

  reply	other threads:[~2020-10-27 22:36 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-14 14:26 [Tarantool-patches] [PATCH 0/2] raft: a few fixes Cyrill Gorcunov
2020-10-14 14:26 ` [Tarantool-patches] [PATCH 1/2] raft: raft_request_to_string -- don't hardcode size Cyrill Gorcunov
2020-10-14 14:26 ` [Tarantool-patches] [PATCH 2/2] raft: decode even invalid states of raft Cyrill Gorcunov
2020-10-15  9:43   ` Serge Petrenko
2020-10-15  9:46     ` Cyrill Gorcunov
2020-10-27 22:18   ` Vladislav Shpilevoy
2020-10-27 22:36     ` Cyrill Gorcunov [this message]
2020-10-15  8:44 ` [Tarantool-patches] [PATCH 0/2] raft: a few fixes Cyrill Gorcunov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201027223630.GC2093@grain \
    --to=gorcunov@gmail.com \
    --cc=tarantool-patches@dev.tarantool.org \
    --cc=v.shpilevoy@tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH 2/2] raft: decode even invalid states of raft' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox