Re: [PATCH v2 4/7] linux/signal.h: Ignore SIGINFO by default in new tasks

[Arseny Maslennikov <ar@cs.msu.ru>]

[-- Attachment #1: Type: text/plain, Size: 2583 bytes --]

On Tue, Jun 25, 2019 at 05:32:15PM -0400, Theodore Ts'o wrote:
> On Tue, Jun 25, 2019 at 07:11:50PM +0300, Arseny Maslennikov wrote:
> > This matches the behaviour of other Unix-like systems that have SIGINFO
> > and causes less harm to processes that do not install handlers for this
> > signal, making the keyboard status character non-fatal for them.
> > 
> > This is implemented with the assumption that SIGINFO is defined
> > to be equivalent to SIGPWR; still, there is no reason for PWR to
> > result in termination of the signal recipient anyway — it does not
> > indicate there is a fatal problem with the recipient's execution
> > context (like e.g. FPE/ILL do), and we have TERM/KILL for explicit
> > termination requests.
> 
> So this is a consequence of trying to overload SIGINFO with SIGPWR.

Pretty much.

> At least on some legacy Unix systems, the way SIGPWR worked was that
> init would broadcast SIGPWR to all userspace process (with system
> daemons on an exception list so they could get shutdown last).
> Applications which didn't have a SIGPWR handler registered, would just
> exit.  Those that did, were expected to clean up in preparation with
> the impending shutdown.  After some period of time, then processes
> would get hard killed and then system daemons would get shut down and
> the system would cleanly shut itself down.
> 
> So SIGPWR acted much like SIGHUP, and that's why the default behavior
> was "terminate".  Now, as far as I know, we've not actually used in
> that way, at least for most common distributions, but there is a sane
> reason why things are they way there are, and in there are people who
> have been using SIGPWR the way it had originally been intended, there
> is risk in overloading SIGINFO with SIGPWR --- in particular, typing
> ^T might cause some enterprise database to start shutting itself down.  :-)

Only if that enterprise database:
1) has a controlling terminal (a strange condition for a daemon, IMHO)
2) that terminal has the status character set in the struct termios.

> 
> (In particular it might be worth checking Linux ports of Oracle and
> DB2.)

A quick google got me this document:
https://docs.oracle.com/cd/B28359_01/server.111/b32009.pdf
It only ever mentions CHLD, CONT, INT, PIPE, TERM, URG and IO.

I have no real experience with neither DB2 nor Oracle DB, though, and no
way to get hold of that kind of software, so I don't know where to look
further. If we'd like to be really sure no one's hurt we'll need someone
with actual expertise here.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]