| 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
| ||
|
Date: Fri, 7 Nov 2008 21:50:34 +0100
From: Lennert Buytenhek <buytenh@...tstofly.org>
To: Bryan Wu <cooloney@...nel.org>
Cc: netdev@...r.kernel.org, Andy Fleming <afleming@...escale.com>,
Byron Bradley <byron.bbradley@...il.com>,
Tim Ellis <tim.ellis@....com>, Imre Kaloz <kaloz@...nwrt.org>,
Nicolas Pitre <nico@....org>, Dirk Teurlings <dirk@...xia.nl>,
Peter van Valderen <p.v.valderen@...il.com>
Subject: Re: [BUG] rmmod bfin_mac netdev driver module will cause system crash.
On Sat, Nov 08, 2008 at 12:40:43AM +0800, Bryan Wu wrote:
> >> >> Hi Lennert,
> >> >>
> >> >> I got a regression since I update our blackfin kernel to the lastest
> >> >> 2.6.28-rc2 which includes your phylib patches.
> >> >> Please find the tracker here:
> >> >> https://blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_id=141&tracker_item_id=4595
> >> >>
> >> >> Briefly, the calling chain is:
> >> >> in bfin_mac_remove(), mdiobus_unregister() ->
> >> >> device_unregister(&bus->dev); -> put_device() -> kobject_put() ->
> >> >> kref_put(&kobj->kref, kobject_release); -> kobject_release() ->
> >> >> kobject_cleanup() -> mdiobus_release().
> >> >> So in mdiobus_unregister() bus->state = MDIOBUS_UNREGISTERED; but in
> >> >> mdiobus_release(), this BUG_ON will fire:
> >> >> BUG_ON(bus->state != MDIOBUS_RELEASED);
> >> >>
> >> >> How about this patch?
> >> >>
> >> >> ---
> >> >> Index: drivers/net/phy/mdio_bus.c
> >> >> ===================================================================
> >> >> --- drivers/net/phy/mdio_bus.c (revision 5575)
> >> >> +++ drivers/net/phy/mdio_bus.c (working copy)
> >> >> @@ -134,13 +134,16 @@
> >> >> int i;
> >> >>
> >> >> BUG_ON(bus->state != MDIOBUS_REGISTERED);
> >> >> - bus->state = MDIOBUS_UNREGISTERED;
> >> >>
> >> >> + bus->state = MDIOBUS_RELEASED;
> >> >> +
> >> >> device_unregister(&bus->dev);
> >> >> for (i = 0; i < PHY_MAX_ADDR; i++) {
> >> >> if (bus->phy_map[i])
> >> >> device_unregister(&bus->phy_map[i]->dev);
> >> >> }
> >> >> +
> >> >> + bus->state = MDIOBUS_UNREGISTERED;
> >> >> }
> >> >> EXPORT_SYMBOL(mdiobus_unregister);
> >> >> ---
> >> >>
> >> >> This regression is gone, but I'm not sure whether there is some side effect.
> >> >> Please review.
> >> >>
> >> >> Thanks a lot
> >> >> -Bryan
> >> >>
> >> >
> >> > Oh, after applying this patch, I met a new WARN_ON issue:
> >> > ---
> >> > root:/> rmmod bfin_mac
> >> > ------------[ cut here ]------------
> >> > WARNING: at lib/kobject.c:593 _kobject_put+0x4e/0x58()
> >> > kobject: '<NULL>' (0098f6a4): is not initialized, yet kobject_put() is
> >> > being called.
> >> > Modules linked in: bfin_mac(-)
> >> > Hardware Trace:
> >> > 0 Target : <0x0000485c> { _dump_stack + 0x0 }
> >> > Source : <0x0000e2fc> { _warn_slowpath + 0x58 } CALL pcrel
> >> > 1 Target : <0x0000e2fc> { _warn_slowpath + 0x58 }
> >> > Source : <0x000294aa> { _print_modules + 0x7a } RTS
> >> > 2 Target : <0x000294a4> { _print_modules + 0x74 }
> >> > Source : <0x0000e9b6> { _printk + 0x16 } RTS
> >> > 3 Target : <0x0000e9b2> { _printk + 0x12 }
> >> > Source : <0x0000f1e8> { _vprintk + 0x130 } RTS
> >> > 4 Target : <0x0000f1c8> { _vprintk + 0x110 }
> >> > Source : <0x0000f30c> { _vprintk + 0x254 } JUMP.S
> >> > 5 Target : <0x0000f30c> { _vprintk + 0x254 }
> >> > Source : <0x0000e826> { _wake_up_klogd + 0x1a } RTS
> >> > 6 Target : <0x0000e826> { _wake_up_klogd + 0x1a }
> >> > Source : <0x0000e818> { _wake_up_klogd + 0xc } IF !CC JUMP
> >> > 7 Target : <0x0000e80c> { _wake_up_klogd + 0x0 }
> >> > Source : <0x0000ec40> { _release_console_sem + 0x1b0 } JUMP.L
> >> > 8 Target : <0x0000ec38> { _release_console_sem + 0x1a8 }
> >> > Source : <0x0000ec2c> { _release_console_sem + 0x19c } IF !CC JUMP
> >> > 9 Target : <0x0000ec28> { _release_console_sem + 0x198 }
> >> > Source : <0x0000ec1a> { _release_console_sem + 0x18a } IF !CC JUMP
> >> > 10 Target : <0x0000ec14> { _release_console_sem + 0x184 }
> >> > Source : <0x00021de8> { _up + 0x3c } RTS
> >> > 11 Target : <0x00021de2> { _up + 0x36 }
> >> > Source : <0x00021dd4> { _up + 0x28 } IF !CC JUMP
> >> > 12 Target : <0x00021dac> { _up + 0x0 }
> >> > Source : <0x0000ec10> { _release_console_sem + 0x180 } CALL pcrel
> >> > 13 Target : <0x0000ebfc> { _release_console_sem + 0x16c }
> >> > Source : <0x0000eae2> { _release_console_sem + 0x52 } IF !CC JUMP
> >> > 14 Target : <0x0000eab2> { _release_console_sem + 0x22 }
> >> > Source : <0x0000ebde> { _release_console_sem + 0x14e } IF !CC JUMP
> >> > 15 Target : <0x0000ebd8> { _release_console_sem + 0x148 }
> >> > Source : <0x0000e776> { __call_console_drivers + 0x7e } RTS
> >> > Stack info:
> >> > SP: [0x00203cfc] <0x00203cfc> /* kernel dynamic memory */
> >> > FP: (0x00203db4)
> >> > Memory from 0x00203cf0 to 00204000
> >> > 00203cf0: ffffff16 00203cfc 001725e4 [0000e2fc]<0000e300> 00203d1c
> >> > 00142380 0016d044
> >> > 00203d10: 00142380 00000251 00203d1c 626f6b5f 7463656a 7475705f
> >> > 3478302b 78302f65
> >> > 00203d30: 00003835 0098e190 00c63fb0 002e9e9c <0000eb62> 0016d068
> >> > 0016d068 0016d040
> >> > 00203d50:<0000ebd8><0000ec14> 000032ea 00203df8 <0008e546> 00162b74
> >> > 0000e9a0 0000001f
> >> > 00203d70: fffffff1 0016d070 <0000f30c> 00162b74 0000e9a0 00162b80
> >> > 00000004 0000000f
> >> > 00203d90: 10624dd3 0097d4c4 00c63fb0 000000d0 000000d0 00203e34
> >> > 00203db0 0000ffff
> >> > 00203db0: 00000000 (00000000)<00089f68><00089f7e> 0017b070 0017b070
> >> > 001725e4 0017b070
> >> > 00203dd0: 00000000 <0009f7fa> 0098f76c <0004da72> 037f5000 00c63f58
> >> > <0009f828> 0098f76c
> >> > 00203df0: 0098f600 0017b008 002e9e9c 00203e20 <00089f16> 0098f6a4
> >> > 00975800 000000ac
> >> > 00203e10: 00000002 00000002 0000002e 00203e2c 00000000 0098f6a4
> >> > <0097cd34> 00975b80
> >> > 00203e30: 00000002 00000001 0000001f <000a141a> 0016c084 0097def4
> >> > 0016c130 00000068
> >> > 00203e50: 00000001 00203e64 <000a14b6> 0016c084 0097def4 <000a08f2>
> >> > 0097e000 0097def4
> >> > 00203e70:<000a08f8> 0097e000 0097def4 00000000 00000880 002e9e9c
> >> > <00109720> 0000001f
> >> > 00203e90:<0002a0e2> 0097e000 00000081 0000004c 001e9b94 <00043122>
> >> > 0017be00 6e696662
> >> > 00203eb0: 63616d5f 00000000 00000000 00000000 00000000 00000000
> >> > 00000000 00000000
> >> > 00203ed0: 00000000 00000000 00000000 00000000 00000000 00000000
> >> > 0097e000 00000880
> >> > 00203ef0: 00203ef4 00000000 <ffa008d4> 00029f80 00000081 00000000
> >> > ffffe000 0017c776
> >> > 00203f10: 00c68de0 0015e23c 00000000 0000002e 00000002 002821ac
> >> > 00008000 00000000
> >> > 00203f30: 00000000 00204000 002821ac 002821ac <0029d3c2><ffa00e3c>
> >> > 02002020 00192d99
> >> > 00203f50: 0028cf9f 00192d92 0028cf96 00000000 0000000f 00000000
> >> > 00000000 00000000
> >> > 00203f70: 00000000 00000000 00000000 00000000 00000000 00000000
> >> > 00000000 00000000
> >> > 00203f90: 00000000 00000000 00000000 00000000 00000000 00000000
> >> > 00000000 002d5f50
> >> > 00203fb0: 002e9fe6 002e9e90 002e9e9c 00000001 002e9f80 0000004c
> >> > 002e0664 002e0664
> >> > 00203fd0: 00000081 00000880 00000000 00000002 0000002e 00000000
> >> > 002e065c 00000880
> >> > 00203ff0: 002e065c 002e065c 00000081 00000006 00000000
> >> > Return addresses in stack:
> >> > address : <0x0000e2fc> { _warn_slowpath + 0x58 }
> >> > address : <0x0000e300> { _warn_slowpath + 0x5c }
> >> > address : <0x0000eb62> { _release_console_sem + 0xd2 }
> >> > address : <0x0000ebd8> { _release_console_sem + 0x148 }
> >> > address : <0x0000ec14> { _release_console_sem + 0x184 }
> >> > address : <0x0008e546> { _vscnprintf + 0x16 }
> >> > address : <0x0000f30c> { _vprintk + 0x254 }
> >> > frame 1 : <0x00089f68> { _kobject_del + 0x10 }
> >> > address : <0x00089f7e> { _kobject_del + 0x26 }
> >> > address : <0x0009f7fa> { _device_del + 0x10a }
> >> > address : <0x0004da72> { _ilookup5 + 0x2e }
> >> > address : <0x0009f828> { _device_unregister + 0xc }
> >> > address : <0x00089f16> { _kobject_put + 0x4e }
> >> > address : <0x0097cd34> { :bfin_mac:_cleanup_module + 0x68 }
> >> > address : <0x000a141a> { ___device_release_driver + 0x5e }
> >> > address : <0x000a14b6> { _driver_detach + 0x72 }
> >> > address : <0x000a08f2> { _bus_remove_driver + 0x6a }
> >> > address : <0x000a08f8> { _bus_remove_driver + 0x70 }
> >> > address : <0x00109720> { _mutex_lock_interruptible + 0xc }
> >> > address : <0x0002a0e2> { _sys_delete_module + 0x162 }
> >> > address : <0x00043122> { _free_bprm + 0x12 }
> >> > address : <0xffa008d4> { _system_call + 0x68 }
> >> > address : <0x0029d3c2> [ rmmod + 0x1d382 ]
> >> > address : <0xffa00e3c> { _evt_system_call + 0x64 }
> >> > ---[ end trace 96c9ba4ee8d1eb27 ]
> >> >
> >> > This WARN was fired by this calling chain: mdiobus_unregister() ->
> >> > put_device() -> kobject_put();
> >> > IMO, mdiobus_release should be called after mdiobus_free, right. So my
> >> > patch is not a good way to solve this bug.
> >> > Any idea about this?
> >> >
> >> > -Bryan
> >> >
> >>
> >> Hi guys,
> >>
> >> I guess I fixed this bug. we need to call
> >> device_unregister(&bus->dev); in mdiobus_free() not in
> >> mdiobus_unregister().
> >> ---
> >> Index: drivers/net/phy/mdio_bus.c
> >> ===================================================================
> >> --- drivers/net/phy/mdio_bus.c (revision 5577)
> >> +++ drivers/net/phy/mdio_bus.c (working copy)
> >> @@ -136,7 +136,6 @@
> >> BUG_ON(bus->state != MDIOBUS_REGISTERED);
> >> bus->state = MDIOBUS_UNREGISTERED;
> >>
> >> - device_unregister(&bus->dev);
> >> for (i = 0; i < PHY_MAX_ADDR; i++) {
> >> if (bus->phy_map[i])
> >> device_unregister(&bus->phy_map[i]->dev);
> >> @@ -165,6 +164,7 @@
> >> BUG_ON(bus->state != MDIOBUS_UNREGISTERED);
> >> bus->state = MDIOBUS_RELEASED;
> >>
> >> + device_unregister(&bus->dev);
> >> put_device(&bus->dev);
> >> }
> >> EXPORT_SYMBOL(mdiobus_free);
> >> ---
> >
> > This would end up doing a double-free on 'bus'.
>
> Yes, device_unregister will call put_device firstly. So how about
> remove the put_device() complete.
>
> ---
> Index: drivers/net/phy/mdio_bus.c
> ===================================================================
> --- drivers/net/phy/mdio_bus.c (revision 5577)
> +++ drivers/net/phy/mdio_bus.c (working copy)
> @@ -136,7 +136,6 @@
> BUG_ON(bus->state != MDIOBUS_REGISTERED);
> bus->state = MDIOBUS_UNREGISTERED;
>
> - device_unregister(&bus->dev);
> for (i = 0; i < PHY_MAX_ADDR; i++) {
> if (bus->phy_map[i])
> device_unregister(&bus->phy_map[i]->dev);
> @@ -165,6 +164,7 @@
> BUG_ON(bus->state != MDIOBUS_UNREGISTERED);
> bus->state = MDIOBUS_RELEASED;
>
> + device_unregister(&bus->dev);
> - put_device(&bus->dev);
> }
> EXPORT_SYMBOL(mdiobus_free);
> ---
The original idea was that re-registering an mdiobus after
registering it would be allowed, which isn't possible if
mdiobus_register() and mdiobus_unregister() aren't symmetric.
--
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
Powered by blists - more mailing lists