Message ID | m1eiis8uc6.fsf@fess.ebiederm.org |
---|---|
State | RFC, archived |
Delegated to: | David Miller |
Headers | show |
On 04/06/2010 08:38 AM, Eric W. Biederman wrote: > > Right now if you recompile the kernel increasing MAXVIFS > to support more VIFS users of the MRT_ADD_VIF and MRT_DEL_VIF > will break because the ABI changed. > > My goal is an API that works with just a recompile of existing > applications, and an ABI that continues to work for old > applications. > > The unused/dead fields at the end of struct mfcctl make this > exercise more difficult than it should be. > > - Rename the existing struct mfcctl mfcctl_old. > - Define a new and larger struct mfcctl that we can detect > by size. > > The new and larger struct mfcctl won't have trailing garbage > fields so we can accept anything of that size or larger, > and simply ignore the entries that are above MAXVIFS. > > My new struct mfcctl is now 128 bytes which is noticeable on > the stack but should still be small enough not to cause problems. > > v2: Rework the support larger arrays so that most/all? existing > applications can simply be recompiled and work with a larger > maximum number of VIFS. If we're going to change the ABI, can we not support an arbitrary number of VIFS instead of just a larger fixed maximum? Thanks, Ben
Ben Greear <greearb@candelatech.com> writes: > On 04/06/2010 08:38 AM, Eric W. Biederman wrote: >> >> Right now if you recompile the kernel increasing MAXVIFS >> to support more VIFS users of the MRT_ADD_VIF and MRT_DEL_VIF >> will break because the ABI changed. >> >> My goal is an API that works with just a recompile of existing >> applications, and an ABI that continues to work for old >> applications. >> >> The unused/dead fields at the end of struct mfcctl make this >> exercise more difficult than it should be. >> >> - Rename the existing struct mfcctl mfcctl_old. >> - Define a new and larger struct mfcctl that we can detect >> by size. >> >> The new and larger struct mfcctl won't have trailing garbage >> fields so we can accept anything of that size or larger, >> and simply ignore the entries that are above MAXVIFS. >> >> My new struct mfcctl is now 128 bytes which is noticeable on >> the stack but should still be small enough not to cause problems. >> >> v2: Rework the support larger arrays so that most/all? existing >> applications can simply be recompiled and work with a larger >> maximum number of VIFS. > > If we're going to change the ABI, can we not support an arbitrary > number of VIFS instead of just a larger fixed maximum? The ABI as I have specified should work for any larger structure than I have specified. But like select many applications will limit themselves to use the definition of struct mfcctl that is passed to them. Eric -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/include/linux/mroute.h b/include/linux/mroute.h index c5f3d53..6dbdebf 100644 --- a/include/linux/mroute.h +++ b/include/linux/mroute.h @@ -76,15 +76,25 @@ struct vifctl { * Cache manipulation structures for mrouted and PIMd */ +struct mfcctl_old { +#define MFCCTL_OLD_VIFS 32 + struct in_addr mfcc_origin; /* Origin of mcast */ + struct in_addr mfcc_mcastgrp; /* Group in question */ + vifi_t mfcc_parent; /* Where it arrived */ + unsigned char mfcc_ttls[MFCCTL_OLD_VIFS]; /* Where it is going */ + unsigned int mfcc_pkt_cnt; /* dead */ + unsigned int mfcc_byte_cnt; /* dead */ + unsigned int mfcc_wrong_if; /* dead */ + int mfcc_expire; /* dead */ +}; + struct mfcctl { +#define MFCCTL_VIFS 118 struct in_addr mfcc_origin; /* Origin of mcast */ struct in_addr mfcc_mcastgrp; /* Group in question */ vifi_t mfcc_parent; /* Where it arrived */ - unsigned char mfcc_ttls[MAXVIFS]; /* Where it is going */ - unsigned int mfcc_pkt_cnt; /* pkt count for src-grp */ - unsigned int mfcc_byte_cnt; - unsigned int mfcc_wrong_if; - int mfcc_expire; + unsigned char mfcc_ttls[MFCCTL_VIFS]; /* Where it is going */ + /* Don't put anything here as mfcc_ttls should grow into here */ }; /* diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c index 0b9d03c..516289b 100644 --- a/net/ipv4/ipmr.c +++ b/net/ipv4/ipmr.c @@ -1012,10 +1012,18 @@ int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsi */ case MRT_ADD_MFC: case MRT_DEL_MFC: - if (optlen != sizeof(mfc)) + + if (optlen == sizeof(struct mfcctl_old)) { + if (copy_from_user(&mfc, optval, sizeof(struct mfcctl_old))) + return -EFAULT; + memset(mfc.mfcc_ttls + MFCCTL_OLD_VIFS, 255, + MFCCTL_VIFS - MFCCTL_OLD_VIFS); + } else if (optlen >= (sizeof(struct mfcctl))) { + if (copy_from_user(&mfc, optval, sizeof(mfc))) + return -EFAULT; + } else return -EINVAL; - if (copy_from_user(&mfc, optval, sizeof(mfc))) - return -EFAULT; + rtnl_lock(); if (optname == MRT_DEL_MFC) ret = ipmr_mfc_delete(net, &mfc); @@ -2032,6 +2040,9 @@ int __init ip_mr_init(void) { int err; + BUILD_BUG_ON(MFCCTL_VIFS < MAXVIFS); + BUILD_BUG_ON(sizeof(struct mfcctl) <= sizeof(struct mfcctl_old)); + mrt_cachep = kmem_cache_create("ip_mrt_cache", sizeof(struct mfc_cache), 0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
Right now if you recompile the kernel increasing MAXVIFS to support more VIFS users of the MRT_ADD_VIF and MRT_DEL_VIF will break because the ABI changed. My goal is an API that works with just a recompile of existing applications, and an ABI that continues to work for old applications. The unused/dead fields at the end of struct mfcctl make this exercise more difficult than it should be. - Rename the existing struct mfcctl mfcctl_old. - Define a new and larger struct mfcctl that we can detect by size. The new and larger struct mfcctl won't have trailing garbage fields so we can accept anything of that size or larger, and simply ignore the entries that are above MAXVIFS. My new struct mfcctl is now 128 bytes which is noticeable on the stack but should still be small enough not to cause problems. v2: Rework the support larger arrays so that most/all? existing applications can simply be recompiled and work with a larger maximum number of VIFS. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> --- include/linux/mroute.h | 20 +++++++++++++++----- net/ipv4/ipmr.c | 17 ++++++++++++++--- 2 files changed, 29 insertions(+), 8 deletions(-)