hcq/MaxScale
hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Allow multiple fatal signals

Browse Source 
As long as the same thread never handles more than one fatal signal,
multiple fatal signals can be processed. This should guarantee that the
stacktrace is printed into the log while guaranteeing that recursion never
takes place if the handling of a fatal signal causes a fatal signal to be
emitted.
This commit is contained in:
Markus Mäkelä 2019-12-30 10:58:01 +02:00
parent b46974f3e7
commit 4f1ae70765
No known key found for this signature in database
GPG Key ID: 72D48FCE664F7B19
1 changed files with 13 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
server/core/gateway.cc
View File

@ -402,11 +402,18 @@ static int signal_set(int sig, void (* handler)(int));
static void sigfatal_handler(int i)
{
// The same signal being handled *now* can occur in another thread (and is often likely).
// By setting the default handler here we will always get a core, but not necessarily
// the backtrace into the log file. This should be overhauled to proper signal handling
// (MXS-599).
signal_set(i, SIG_DFL);
thread_local std::thread::id current_id;
std::thread::id no_id;
if (current_id != no_id)
{
// Fatal error when processing a fatal error.
// TODO: This should be overhauled to proper signal handling (MXS-599).
signal_set(i, SIG_DFL);
raise(i);
}
current_id = std::this_thread::get_id();
MXS_CONFIG* cnf = config_get_global_options();
fprintf(stderr,
@ -444,6 +451,7 @@ static void sigfatal_handler(int i)
/* re-raise signal to enforce core dump */
fprintf(stderr, "\n\nWriting core dump\n");
signal_set(i, SIG_DFL);
raise(i);
}