Re: [syzbot] [net?] KASAN: invalid-access Read in __packet_get_status

[Willem de Bruijn]

On Mon, May 22, 2023 at 10:52 AM Willem de Bruijn
<willemdebruijn.kernel@xxxxxxxxx> wrote:
>
> On Mon, May 22, 2023 at 6:51 AM syzbot
> <syzbot+64b0f633159fde08e1f1@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit:    2d1bcbc6cd70 Merge tag 'probes-fixes-v6.4-rc1' of git://gi..
> > git tree:       upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=154b8fa1280000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=51dd28037b2a55f
> > dashboard link: https://syzkaller.appspot.com/bug?extid=64b0f633159fde08e1f1
> > compiler:       aarch64-linux-gnu-gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
> > userspace arch: arm64
> > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=12b6382e280000
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=17fd0aee280000
> >
> > Downloadable assets:
> > disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/384ffdcca292/non_bootable_disk-2d1bcbc6.raw.xz
> > vmlinux: https://storage.googleapis.com/syzbot-assets/d2e21a43e11e/vmlinux-2d1bcbc6.xz
> > kernel image: https://storage.googleapis.com/syzbot-assets/49e0b029f9af/Image-2d1bcbc6.gz.xz
> >
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+64b0f633159fde08e1f1@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> > ==================================================================
> > BUG: KASAN: invalid-access in __packet_get_status+0x70/0xe0 net/packet/af_packet.c:438
>
> The offending line is the last one in
>
> "
> static int __packet_get_status(const struct packet_sock *po, void *frame)
> {
>         union tpacket_uhdr h;
>
>         smp_rmb();
>
>         h.raw = frame;
>         switch (po->tp_version) {
>         case TPACKET_V1:
>                 flush_dcache_page(pgv_to_page(&h.h1->tp_status));
>                 return h.h1->tp_status;
>         case TPACKET_V2:
>                 flush_dcache_page(pgv_to_page(&h.h2->tp_status));
> "
>
> The reproducer is very small:
>
> "
> // socket(PF_PACKET, SOCK_DGRAM, htons(ETH_P_ALL);
> r0 = socket$packet(0x11, 0x2, 0x300)
>
> // setsockopt PACKET_RX_RING with same block and frame sizes and counts
> setsockopt$packet_rx_ring(r0, 0x107, 0x5,
> &(0x7f0000000040)=@req3={0x8000, 0x200, 0x80, 0x20000}, 0x1c)
>
> // excessive length, too many bits in prot, MAP_SHARED | MAP_ANONYMOUS
> mmap(&(0x7f0000568000/0x2000)=nil, 0x1000000, 0x20567fff, 0x11, r0, 0x0)
> "
>
> What is odd here is that the program never sets packet version
> explicitly, and the default is TPACKET_V1.

The test is marked as repeat.

One possibility is that there is a race between packet arrival calling
flush_dcache_page and user mmap setup/teardown. That would exhibit as
flakiness.

ARM flush_dcache_page is quite outside my networking comfort zone.