Tarantool development patches archive
 help / color / mirror / Atom feed
From: Georgy Kirichenko <georgy@tarantool.org>
To: tarantool-patches@freelists.org
Cc: Georgy Kirichenko <georgy@tarantool.org>
Subject: [tarantool-patches] [PATCH 2/3] Proper unwind for currently executing fiber
Date: Thu,  4 Oct 2018 10:44:16 +0300	[thread overview]
Message-ID: <3bf7292e6ae0c5b7b892c204457c630c73bd8108.1538638419.git.georgy@tarantool.org> (raw)
In-Reply-To: <20180926173625.GO30528@chai>

Each yielded fiber has a preserved coro state stored in a corresponding
variable however an executing fiber has a volatile state placed in CPU
registers (stack pointer, instruction pointer and non-volatile registers)
and corresponding context-storing variable value is invalid.
For already yielded fiber we have to use a special asm-written handler to make
a temporary switch to the preserved state and capture executing context what
is not needed for executing fiber.
After the patch for the executing fiber NULL is passed to the backtrace
function as coro context and then backtrace function could decide should
it use special context-switching handler or might just use unw_getcontext
from the unwind library.
---
 src/backtrace.cc | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

diff --git a/src/backtrace.cc b/src/backtrace.cc
index 2512bc045..d59fd5879 100644
--- a/src/backtrace.cc
+++ b/src/backtrace.cc
@@ -363,12 +363,32 @@ __asm__ volatile(
 #endif
 }
 
+/*
+ * Call `cb' callback for each `coro_ctx' contained frame or the current
+ * executed coroutine if `coro_ctx' is NULL. A coro_context is a structure
+ * created on each coroutine yield to store execution context so for an on-CPU
+ * coroutine there is no valid coro_context could be defined and NULL is
+ * passed.
+*/
 void
 backtrace_foreach(backtrace_cb cb, coro_context *coro_ctx, void *cb_ctx)
 {
 	unw_cursor_t unw_cur;
 	unw_context_t unw_ctx;
-	coro_unwcontext(&unw_ctx, coro_ctx);
+	if (coro_ctx == NULL) {
+		/*
+		 * Current executing coroutine and simple unw_getcontext
+		 * should function.
+		 */
+		unw_getcontext(&unw_ctx);
+	} else {
+		/*
+		 * Execution context is stored in the coro_ctx
+		 * so use special context-switching handler to
+		 * capture an unwind context.
+		 */
+		coro_unwcontext(&unw_ctx, coro_ctx);
+	}
 	unw_init_local(&unw_cur, &unw_ctx);
 	int frame_no = 0;
 	unw_word_t sp = 0, old_sp = 0, ip, offset;
-- 
2.19.0

  reply	other threads:[~2018-10-04  7:44 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-09-21 13:20 [tarantool-patches] [PATCH 0/3] Dump lua frames for a fiber traceback Georgy Kirichenko
2018-09-21 13:20 ` [tarantool-patches] [PATCH 1/3] Set a lua state for the main fiber too Georgy Kirichenko
2018-09-25 23:27   ` [tarantool-patches] " Konstantin Osipov
2018-10-05 15:04   ` [tarantool-patches] " Vladimir Davydov
2018-09-21 13:20 ` [tarantool-patches] [PATCH 2/3] Proper unwind for currently executing fiber Georgy Kirichenko
2018-09-25 23:34   ` [tarantool-patches] " Konstantin Osipov
2018-09-26  6:57     ` Georgy Kirichenko
2018-09-26 17:36       ` Konstantin Osipov
2018-10-04  7:44         ` Georgy Kirichenko [this message]
2018-10-04 22:27           ` Konstantin Osipov
2018-10-05 15:09           ` [tarantool-patches] " Vladimir Davydov
2018-09-21 13:20 ` [tarantool-patches] [PATCH 3/3] Show names of Lua functions in backtraces Georgy Kirichenko
2018-09-25 23:38   ` [tarantool-patches] " Konstantin Osipov
2018-10-04  7:55     ` [tarantool-patches] " Georgy Kirichenko
2018-10-04 22:28       ` [tarantool-patches] " Konstantin Osipov
2018-10-05 17:49       ` [tarantool-patches] " Vladimir Davydov
2018-10-10 16:39 [tarantool-patches] [PATCH v2 0/3] Dump lua frames for a fiber traceback Georgy Kirichenko
2018-10-10 16:39 ` [tarantool-patches] [PATCH 2/3] Proper unwind for currently executing fiber Georgy Kirichenko

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=3bf7292e6ae0c5b7b892c204457c630c73bd8108.1538638419.git.georgy@tarantool.org \
    --to=georgy@tarantool.org \
    --cc=tarantool-patches@freelists.org \
    --subject='Re: [tarantool-patches] [PATCH 2/3] Proper unwind for currently executing fiber' \
    /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