netdev - [PATCH] e100 module loads 1/2 times

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20070813075453.GA4257@1wt.eu>
Date:	Mon, 13 Aug 2007 09:54:53 +0200
From:	Willy Tarreau <w@....eu>
To:	uke Kok <auke-jan.h.kok@...el.com>,
	"Brandeburg, Jesse" <jesse.brandeburg@...el.com>
Cc:	netdev@...r.kernel.org
Subject: [PATCH] e100 module loads 1/2 times

Hi Auke, Jesse,

for a long time, I've been annoyed by version 3.5.17 of the e100 driver
which refuses to load on first time and only loads on second time. Since
I always had the original 2.3.43 driver in kernel 2.4, I did not care
that much. Recently, I encountered real troubles with 2.3.43 in a 802.1q
setup (basically it did not untag incoming frames). So I decided to give
3.5.17 a second try. Same problem, I had to load it twice.

I finally found the problem in e100_init_module(). Up to 3.5.14, the
return of pci_module_init() was returned. This one equals zero if
everything went fine, <0 otherwise, which is compatible with init_module().

With 3.5.17, the result comes from pci_register_driver(), which returns
the number of devices registered. So the problem now makes sense :

- first call: the driver registers itself and returns non-zero, which
  is an error for insmod

- second call: the driver cannot register again and returns zero new
  drivers, which is good for insmod.

Note that e1000 has a related bug : it uses the return from pci_module_init()
to decide whether or not to register a reboot notifier. Fortunately, the
test is performed with ret>=0, which matches ==0 and not <0 (errors), so
this works as a side effect.

The obvious fix for e100 reusing pci_module_init is below. Also, since
e100-2.3.43 does not work at all with vlans in 2.4, I was thinking about
upgrading it to 3.5.17. It would also be the same version as in 2.6,
simplifying its long-term maintenance. What do you think about this ?

Best regards,
Willy

--- e100-3.5.17/src/e100.c.orig	2007-08-13 08:53:18 +0200
+++ e100-3.5.17/src/e100.c	2007-08-13 09:24:56 +0200
@@ -2934,13 +2934,13 @@
 		printk(KERN_INFO PFX "%s\n", DRV_COPYRIGHT);
 	}
 #ifdef E100_USE_REBOOT_NOTIFIER
-	retval = pci_register_driver(&e100_driver);
-	if (retval >= 0)
+	retval = pci_module_init(&e100_driver);
+	if (retval == 0)
 		register_reboot_notifier(&e100_notifier_reboot);

 	return retval;
 #else
-	return pci_register_driver(&e100_driver);
+	return pci_module_init(&e100_driver);
 #endif
 }

-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html