Re: [PATCH net] net: tun: Update napi->skb after XDP process

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Willem de Bruijn wrote:
> Wang Liang wrote:
> > The syzbot report a UAF issue:
> > 
> >   BUG: KASAN: slab-use-after-free in skb_reset_mac_header include/linux/skbuff.h:3150 [inline]
> >   BUG: KASAN: slab-use-after-free in napi_frags_skb net/core/gro.c:723 [inline]
> >   BUG: KASAN: slab-use-after-free in napi_gro_frags+0x6e/0x1030 net/core/gro.c:758
> >   Read of size 8 at addr ffff88802ef22c18 by task syz.0.17/6079
> >   CPU: 0 UID: 0 PID: 6079 Comm: syz.0.17 Not tainted syzkaller #0 PREEMPT(full)
> >   Call Trace:
> >    <TASK>
> >    dump_stack_lvl+0x189/0x250 lib/dump_stack.c:120
> >    print_address_description mm/kasan/report.c:378 [inline]
> >    print_report+0xca/0x240 mm/kasan/report.c:482
> >    kasan_report+0x118/0x150 mm/kasan/report.c:595
> >    skb_reset_mac_header include/linux/skbuff.h:3150 [inline]
> >    napi_frags_skb net/core/gro.c:723 [inline]
> >    napi_gro_frags+0x6e/0x1030 net/core/gro.c:758
> >    tun_get_user+0x28cb/0x3e20 drivers/net/tun.c:1920
> >    tun_chr_write_iter+0x113/0x200 drivers/net/tun.c:1996
> >    new_sync_write fs/read_write.c:593 [inline]
> >    vfs_write+0x5c9/0xb30 fs/read_write.c:686
> >    ksys_write+0x145/0x250 fs/read_write.c:738
> >    do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> >    do_syscall_64+0xfa/0x3b0 arch/x86/entry/syscall_64.c:94
> >    entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >    </TASK>
> > 
> >   Allocated by task 6079:
> >    kasan_save_stack mm/kasan/common.c:47 [inline]
> >    kasan_save_track+0x3e/0x80 mm/kasan/common.c:68
> >    unpoison_slab_object mm/kasan/common.c:330 [inline]
> >    __kasan_mempool_unpoison_object+0xa0/0x170 mm/kasan/common.c:558
> >    kasan_mempool_unpoison_object include/linux/kasan.h:388 [inline]
> >    napi_skb_cache_get+0x37b/0x6d0 net/core/skbuff.c:295
> >    __alloc_skb+0x11e/0x2d0 net/core/skbuff.c:657
> >    napi_alloc_skb+0x84/0x7d0 net/core/skbuff.c:811
> >    napi_get_frags+0x69/0x140 net/core/gro.c:673
> >    tun_napi_alloc_frags drivers/net/tun.c:1404 [inline]
> >    tun_get_user+0x77c/0x3e20 drivers/net/tun.c:1784
> >    tun_chr_write_iter+0x113/0x200 drivers/net/tun.c:1996
> >    new_sync_write fs/read_write.c:593 [inline]
> >    vfs_write+0x5c9/0xb30 fs/read_write.c:686
> >    ksys_write+0x145/0x250 fs/read_write.c:738
> >    do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> >    do_syscall_64+0xfa/0x3b0 arch/x86/entry/syscall_64.c:94
> >    entry_SYSCALL_64_after_hwframe+0x77/0x7f
> > 
> >   Freed by task 6079:
> >    kasan_save_stack mm/kasan/common.c:47 [inline]
> >    kasan_save_track+0x3e/0x80 mm/kasan/common.c:68
> >    kasan_save_free_info+0x46/0x50 mm/kasan/generic.c:576
> >    poison_slab_object mm/kasan/common.c:243 [inline]
> >    __kasan_slab_free+0x5b/0x80 mm/kasan/common.c:275
> >    kasan_slab_free include/linux/kasan.h:233 [inline]
> >    slab_free_hook mm/slub.c:2422 [inline]
> >    slab_free mm/slub.c:4695 [inline]
> >    kmem_cache_free+0x18f/0x400 mm/slub.c:4797
> >    skb_pp_cow_data+0xdd8/0x13e0 net/core/skbuff.c:969
> >    netif_skb_check_for_xdp net/core/dev.c:5390 [inline]
> >    netif_receive_generic_xdp net/core/dev.c:5431 [inline]
> >    do_xdp_generic+0x699/0x11a0 net/core/dev.c:5499
> >    tun_get_user+0x2523/0x3e20 drivers/net/tun.c:1872
> >    tun_chr_write_iter+0x113/0x200 drivers/net/tun.c:1996
> >    new_sync_write fs/read_write.c:593 [inline]
> >    vfs_write+0x5c9/0xb30 fs/read_write.c:686
> >    ksys_write+0x145/0x250 fs/read_write.c:738
> >    do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> >    do_syscall_64+0xfa/0x3b0 arch/x86/entry/syscall_64.c:94
> >    entry_SYSCALL_64_after_hwframe+0x77/0x7f
> > 
> > After commit e6d5dbdd20aa ("xdp: add multi-buff support for xdp running in
> > generic mode"), the original skb may be freed in skb_pp_cow_data() when
> > XDP program was attached, which was allocated in tun_napi_alloc_frags().
> > However, the napi->skb still point to the original skb, update it after
> > XDP process.
> > 
> > Reported-by: syzbot+64e24275ad95a915a313@syzkaller.appspotmail.com
> > Closes: https://syzkaller.appspot.com/bug?extid=64e24275ad95a915a313
> > Fixes: e6d5dbdd20aa ("xdp: add multi-buff support for xdp running in generic mode")
> > Signed-off-by: Wang Liang <wangliang74@huawei.com>
> > ---
> >  drivers/net/tun.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> > index cc6c50180663..47ddcb4b9a78 100644
> > --- a/drivers/net/tun.c
> > +++ b/drivers/net/tun.c
> > @@ -1875,6 +1875,9 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
> >  				local_bh_enable();
> >  				goto unlock_frags;
> >  			}
> > +
> > +			if (frags && skb != tfile->napi.skb)
> > +				tfile->napi.skb = skb;
> 
> This is observed with tun because syzkaller can fuzz napi with that.
> That unfortunately added fuzz test coverage to a combination that is
> not intended for real use: XDP generic before napi frags.
> 
> Tun is the only driver that calls do_xdp_generic on a napi.skb and
> later passes this napi to napi_gro_frags.
> 
> But this is no longer a napi frags skb on which napi_gro_frags
> (and napi_frags_skb and gro_pull_from_frag0) should be called? As the
> skb now has a linear part. Not sure that the frag0 is still correct.

Never mind this. napi_alloc_skb may also fall back to a regular
__skb_alloc and pass that a napi.skb to napi_gro_frags. Which
reinitializes the NAPI_GRO_CB and with that frag0 accordingly.