On 22.03, Pablo Neira Ayuso wrote: > On Sun, Mar 22, 2015 at 10:05:10AM +0200, Denys Fedoryshchenko wrote: > > On 2015-03-22 07:33, Patrick McHardy wrote: > > >On 22.03, Denys Fedoryshchenko wrote: > > >>Sorry for noise, seems git version working fine! > > > > > >Still this shouldn't be happening. Just to confirm, you were using an > > >unpatched kernel and by git you mean nftables git? > > > > Yes, correct. I tested on 3.18.8 and 3.19.2 vanilla kernels (x86_64). > > On nftables 0.4 it does crash, on nftables git it doesn't. > > I sent this fix to -stable by March 10th but this doesn't show up in > 3.18.x and 3.19.x yet. > > [ upstream commit 02263db00b6cb98701332aa257c07ca549c2324b ] I think this is actually a different problem. We're using set->dtype for uninit of the element's data, but unless it's NFT_DATA_VERDICT, its holding the user encoding of the type. Basically all the types except NFT_DATA_RESERVED_MASK map to NFT_DATA_VALUE, and it seems we're not properly handling it in that path. > > We have several problems in this path: > > 1) There is a use-after-free when removing individual elements from > the commit path. > > 2) We have to uninit() the data part of the element from the abort > path to avoid a chain refcount leak. > > 3) We have to check for set->flags to see if there's a mapping, > instead > of the element flags. > > 4) We have to check for !(flags & NFT_SET_ELEM_INTERVAL_END) to skip > elements that are part of the interval that have no data part, so > they don't need to be uninit(). > > Cc: <stable@xxxxxxxxxxxxxxx> # 3.18.x > Cc: <stable@xxxxxxxxxxxxxxx> # 3.19.x > Signed-off-by: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> > > > >>On 2015-03-22 00:49, Denys Fedoryshchenko wrote: > > >>>Additionally, if i will do "nft flush table mangle" , with this table > > >>>added i will get this: > > >>>[ 42.800078] ------------[ cut here ]------------ > > >>>[ 42.800092] WARNING: CPU: 3 PID: 2868 at > > >>>net/netfilter/nf_tables_api.c:4122 nft_data_uninit+0x35/0x50 > > >>>[nf_tables]() > -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html