net: restore ip source validation
in [Kernel]
|
Prev: [PATCH] driver-core: fix race condition in get_device_parent()
Next: Inconsistency in ioread functions, const pointer or not
From: Sven Joachim on 5 Feb 2010 16:00 On 2010-02-05 21:17 +0100, Eric W. Biederman wrote: > Bah. That DEVINET_SYSCTL_ENTRY requires having a binary sysctl > assigned, just to use as an index. > > Which of course trips over all of the fine checks in sysctl_check.c to > keep people from assigning new binary sysctls by accident. > > That is the only place in the kernel where he have that problem, I wonder > how much work it will be to finish untangling. Isn't that already done in 2.6.33, looking at commit 83ac201b ? Forgive my ignorance, I am a layman. Sen -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Eric W. Biederman on 5 Feb 2010 17:30 Sven Joachim <svenjoac(a)gmx.de> writes: > On 2010-02-05 21:17 +0100, Eric W. Biederman wrote: > >> Bah. That DEVINET_SYSCTL_ENTRY requires having a binary sysctl >> assigned, just to use as an index. >> >> Which of course trips over all of the fine checks in sysctl_check.c to >> keep people from assigning new binary sysctls by accident. >> >> That is the only place in the kernel where he have that problem, I wonder >> how much work it will be to finish untangling. > > Isn't that already done in 2.6.33, looking at commit 83ac201b ? Forgive > my ignorance, I am a layman. In 2.6.33 the enumeration in sysctl.h still serves double duty as an index into a per network device bitmap and as the binary sysctl number. You are correct that the rest of the binary sysctl code is decoupled in 2.6.33. In 2.6.32 the implementation is also still coupled and that difference is what caused problems for the backport. I just wince whenever I noticed we have touched sysctl.h. Eric -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Eric W. Biederman on 5 Feb 2010 17:30
Sven Joachim <svenjoac(a)gmx.de> writes: > On 2010-02-05 21:17 +0100, Eric W. Biederman wrote: > >> Bah. That DEVINET_SYSCTL_ENTRY requires having a binary sysctl >> assigned, just to use as an index. >> >> Which of course trips over all of the fine checks in sysctl_check.c to >> keep people from assigning new binary sysctls by accident. >> >> That is the only place in the kernel where he have that problem, I wonder >> how much work it will be to finish untangling. > > Isn't that already done in 2.6.33, looking at commit 83ac201b ? Forgive > my ignorance, I am a layman. In 2.6.33 the enumeration in sysctl.h still serves double duty as an index into a per network device bitmap and as the binary sysctl number. You are correct that the rest of the binary sysctl code is decoupled in 2.6.33. In 2.6.32 the implementation is also still coupled and that difference is what caused problems for the backport. I just wince whenever I noticed we have touched sysctl.h. Eric -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |