Skip to contentSkip to conversation

metag test failure in -next due to commit 'module: relocate module_init from init.h to module.h'

DISCUSSION Guenter Roeck 3 participants 4 messages Jun 5, 2015 – Jun 7, 2015 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

Hi,

My qemu metag runtime test fails in -next due to commit 5ebbb3badaa8 ("module:
relocate module_init from init.h to module.h"). Bisect log is as follows. A log of a failed test is available at
http://server.roeck-us.net:8010/builders/qemu-metag-next/builds/173/steps/qemubuildcommand/logs/stdio

Essentially the test just hangs in boot.

Reverting the commit fixes the problem.

Guenter
---

# bad: [94feb25d6fcd34b0d45c18cf0185d83d0c13c60a] Add linux-next specific files for 20150604
# good: [c65b99f046843d2455aa231747b5a07a999a9f3d] Linux 4.1-rc6
git bisect start 'HEAD' 'v4.1-rc6'
# good: [46cb2717b4b827977bccd276096791d0efdb4119] Merge remote-tracking branch 'drm/drm-next'
git bisect good 46cb2717b4b827977bccd276096791d0efdb4119
# good: [37104644ae9384d6fc6574f16dc9782259ab89b1] Merge remote-tracking branch 'kvms390/next'
git bisect good 37104644ae9384d6fc6574f16dc9782259ab89b1
# good: [311942c925377a65157e4b14778a22bb05808314] Merge remote-tracking branch 'staging/staging-next' git bisect good 311942c925377a65157e4b14778a22bb05808314
# good: [4b40a87efb321750a527aea556337f24b5268783] Merge remote-tracking branch 'clk/clk-next'
git bisect good 4b40a87efb321750a527aea556337f24b5268783
# good: [0cd6ba7205ab368b324d5a5f601aa3f5f89c246f] MAINTAINERS: update sound soc intel patterns
git bisect good 0cd6ba7205ab368b324d5a5f601aa3f5f89c246f
# good: [076edbd2f016f2fb4a80b95625cd65a9aed4e06c] Merge remote-tracking branch 'rtc/rtc-next'
git bisect good 076edbd2f016f2fb4a80b95625cd65a9aed4e06c
# good: [8c169d94c31fced71d907e26b19b4f5f0bd9d7bf] scsi-resolve-sg-buffer-const-ness-issue-fix
git bisect good 8c169d94c31fced71d907e26b19b4f5f0bd9d7bf
# good: [baaf163cd06a517d85dc39a6fa00107e3620b779] Merge branch 'module_init-device_initcall-4.1-rc6' into init-v4.1-rc6 git bisect good baaf163cd06a517d85dc39a6fa00107e3620b779
# good: [8f0b9591c464a9fd783b481b2c9cddf008043ba4] Merge branch 'module-builtin_driver-4.1-rc6' into init-v4.1-rc6 git bisect good 8f0b9591c464a9fd783b481b2c9cddf008043ba4
# bad: [fc3af372b8c176c08e8921d772ca2bff5620f15b] kernel/relay.c: use kvfree() in relay_free_page_array() git bisect bad fc3af372b8c176c08e8921d772ca2bff5620f15b
# bad: [cdcdbb925bb730624d1312620263c0cf3666da5c] Merge remote-tracking branch 'init/init-v4.1-rc6' git bisect bad cdcdbb925bb730624d1312620263c0cf3666da5c
# good: [3a283d1794f58a054f247152fb4aa15088aad561] mips: make loongsoon serial driver explicitly modular git bisect good 3a283d1794f58a054f247152fb4aa15088aad561
# bad: [3491bfd125cd3d2b7654e2244fb1c262c7212049] Merge branch 'module-misc-4.1-rc6' into init-v4.1-rc6 git bisect bad 3491bfd125cd3d2b7654e2244fb1c262c7212049
# bad: [5ebbb3badaa8ffc50558af88cdea27497953356a] module: relocate module_init from init.h to module.h git bisect bad 5ebbb3badaa8ffc50558af88cdea27497953356a

# first bad commit: [5ebbb3badaa8ffc50558af88cdea27497953356a] module: relocate module_init from init.h to module.h
#1 lore raw
Quoted 8 lines · Guenter Roeck

My qemu metag runtime test fails in -next due to commit 5ebbb3badaa8 ("module:
relocate module_init from init.h to module.h"). Bisect log is as follows. A log of a failed test is available at
http://server.roeck-us.net:8010/builders/qemu-metag-next/builds/173/steps/qemubuildcommand/logs/stdio

Essentially the test just hangs in boot.

Reverting the commit fixes the problem.

Thanks for finding and bisecting Guenter!

Some build warnings from the same commit caught my eye, which made it a quick one to figure out. Paul, please consider applying the patch below before the offending commit "module: relocate module_init from init.h to module.h", so as not to break bisection.

Cheers
James

From e12856c559d7dff2ad4f6497996610e12e7c7e2d Mon Sep 17 00:00:00 2001
From: James Hogan <[email protected]>
Date: Fri, 5 Jun 2015 22:17:18 +0100
Subject: [PATCH] tty/metag_da: Avoid module_init/module_exit
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The metag_da TTY driver can't get built as a module at the moment, but it still uses module_init() and module_exit(). Those macros are moving to module.h which isn't included by metag_da.c, which will result in the following build warnings (remarkably no build errors) and an apparent failure to boot as the TTY driver won't be loaded.

drivers/tty/metag_da.c:660: warning: data definition has no type or storage class drivers/tty/metag_da.c:660: warning: type defaults to ‘int’ in declaration of ‘module_init’ drivers/tty/metag_da.c:660: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:661: warning: data definition has no type or storage class drivers/tty/metag_da.c:661: warning: type defaults to ‘int’ in declaration of ‘module_exit’ drivers/tty/metag_da.c:661: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:572: warning: ‘dashtty_init’ defined but not used drivers/tty/metag_da.c:645: warning: ‘dashtty_exit’ defined but not used
drivers/tty/metag_da.c In function ‘dash_console_write’:
drivers/tty/metag_da.c:670 : warning: passing argument 4 of ‘chancall’ discards qualifiers from pointer target type

Instead of just adding the module.h include, now would be a good time to remove the use of these macros, replacing the module_init with device_initcall, and removing the exit function altogether since it isn't needed. If module support is added later the code can always be resurrected.

---
 drivers/tty/metag_da.c | 20 +-------------------
 1 file changed, 1 insertion(+), 19 deletions(-)
drivers/tty/metag_da.c +1 -19
index 3774600741d8..9325262289f9 100644
@@ -640,25 +640,7 @@ err_destroy_ports:
 	put_tty_driver(channel_driver);
 	return ret;
 }
-
-static void dashtty_exit(void)
-{
-	int nport;
-	struct dashtty_port *dport;
-
-	del_timer_sync(&put_timer);
-	kthread_stop(dashtty_thread);
-	del_timer_sync(&poll_timer);
-	tty_unregister_driver(channel_driver);
-	for (nport = 0; nport < NUM_TTY_CHANNELS; nport++) {
-		dport = &dashtty_ports[nport];
-		tty_port_destroy(&dport->port);
-	}
-	put_tty_driver(channel_driver);
-}
-
-module_init(dashtty_init);
-module_exit(dashtty_exit);
+device_initcall(dashtty_init);
 
 #ifdef CONFIG_DA_CONSOLE
 
Signature · 1 line
2.3.6
author Tested-by
#2 lore raw

On 06/05/2015 02:39 PM, James Hogan wrote:
[ ... ]

Quoted 32 lines

From e12856c559d7dff2ad4f6497996610e12e7c7e2d Mon Sep 17 00:00:00 2001
From: James Hogan <[email protected]>
Date: Fri, 5 Jun 2015 22:17:18 +0100
Subject: [PATCH] tty/metag_da: Avoid module_init/module_exit
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The metag_da TTY driver can't get built as a module at the moment, but it still uses module_init() and module_exit(). Those macros are moving to module.h which isn't included by metag_da.c, which will result in the following build warnings (remarkably no build errors) and an apparent failure to boot as the TTY driver won't be loaded.

drivers/tty/metag_da.c:660: warning: data definition has no type or storage class drivers/tty/metag_da.c:660: warning: type defaults to ‘int’ in declaration of ‘module_init’ drivers/tty/metag_da.c:660: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:661: warning: data definition has no type or storage class drivers/tty/metag_da.c:661: warning: type defaults to ‘int’ in declaration of ‘module_exit’ drivers/tty/metag_da.c:661: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:572: warning: ‘dashtty_init’ defined but not used drivers/tty/metag_da.c:645: warning: ‘dashtty_exit’ defined but not used
drivers/tty/metag_da.c In function ‘dash_console_write’:
drivers/tty/metag_da.c:670 : warning: passing argument 4 of ‘chancall’ discards qualifiers from pointer target type

Instead of just adding the module.h include, now would be a good time to remove the use of these macros, replacing the module_init with device_initcall, and removing the exit function altogether since it isn't needed. If module support is added later the code can always be resurrected.

Quoted 6 lines

Yes, that does the trick.

Thanks,
Guenter

#3 lore raw
Quoted 41 lines · [Re: metag test failure in -next due to commit 'module: relocate module_init…

On 06/05/2015 02:39 PM, James Hogan wrote:
[ ... ]

Quoted 32 lines

From e12856c559d7dff2ad4f6497996610e12e7c7e2d Mon Sep 17 00:00:00 2001
From: James Hogan <[email protected]>
Date: Fri, 5 Jun 2015 22:17:18 +0100
Subject: [PATCH] tty/metag_da: Avoid module_init/module_exit
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The metag_da TTY driver can't get built as a module at the moment, but it still uses module_init() and module_exit(). Those macros are moving to module.h which isn't included by metag_da.c, which will result in the following build warnings (remarkably no build errors) and an apparent failure to boot as the TTY driver won't be loaded.

drivers/tty/metag_da.c:660: warning: data definition has no type or storage class drivers/tty/metag_da.c:660: warning: type defaults to ‘int’ in declaration of ‘module_init’ drivers/tty/metag_da.c:660: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:661: warning: data definition has no type or storage class drivers/tty/metag_da.c:661: warning: type defaults to ‘int’ in declaration of ‘module_exit’ drivers/tty/metag_da.c:661: warning: parameter names (without types) in function declaration drivers/tty/metag_da.c:572: warning: ‘dashtty_init’ defined but not used drivers/tty/metag_da.c:645: warning: ‘dashtty_exit’ defined but not used
drivers/tty/metag_da.c In function ‘dash_console_write’:
drivers/tty/metag_da.c:670 : warning: passing argument 4 of ‘chancall’ discards qualifiers from pointer target type

Instead of just adding the module.h include, now would be a good time to remove the use of these macros, replacing the module_init with device_initcall, and removing the exit function altogether since it isn't needed. If module support is added later the code can always be resurrected.

Quoted 6 lines
Quoted 3 lines

Yes, that does the trick.

Thanks guys, I'll add that to the module_init-device_initcall conversion branch, which will put it ahead of the code relocation commit.

I think most other arch run with -Werror-implicit (or whatever it is) so that these kinds of things would not "leak" past into a silent boot fail like this.

Paul.
--

Quoted 4 lines

Thanks,
Guenter