Content-Type: multipart/mixed; boundary="===============1920542019371304447==" MIME-Version: 1.0 From: Denis Kenzior Subject: Re: [PATCH] monitor: Add message length check to nlmon_receice Date: Tue, 21 Jan 2020 16:36:37 -0600 Message-ID: <8648726d-dd94-84a6-4d98-7dec428aa55a@gmail.com> In-Reply-To: <20200119163111.31753-1-wagi@monom.org> List-Id: To: iwd@lists.01.org --===============1920542019371304447== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Hi Daniel, On 1/19/20 10:31 AM, Daniel Wagner wrote: > The NLMSG_NEXT macro calculates the next nlmsg and updates the len > field: > = > #define NLMSG_NEXT(nlh,len) ((len) -=3D NLMSG_ALIGN((nlh)->nlmsg_len)= , \ > (struct nlmsghdr*)(((char*)(nlh)) + NLMSG_ALIGN((nlh)->nlmsg_len))) > = > That means nlmsg_len needs to be an multiple of NLMSG_ALIGNTO to avoid > an underflow in len. But there are message which do not have a valid > lenght: > = hmm, do you have a pcap of such a message? > Breakpoint 1, nlmon_receive (io=3D0x4da210, user_data=3D0x4d4cc0) at m= onitor/nlmon.c:6947 > 6947 printf("malformed packet\n"); > (gdb) bt > #0 nlmon_receive (io=3D0x4da210, user_data=3D0x4d4cc0) at monitor/nlm= on.c:6947 > #1 0x0000000000439a91 in io_callback (fd=3D7, events=3D1, user_data= =3D0x4da210) at ell/io.c:126 > #2 0x0000000000438861 in l_main_iterate (timeout=3D-1) at ell/main.c:= 473 > #3 0x0000000000438968 in l_main_run () at ell/main.c:520 > #4 0x0000000000438c80 in l_main_run_with_signal (callback=3D0x4038e6 = , user_data=3D0x0) at ell/main.c:642 > #5 0x0000000000403cc9 in main (argc=3D3, argv=3D0x7fffffffec28) at mo= nitor/main.c:806 > (gdb) p nlmsg->nlmsg_len > $5 =3D 17 > = > By adding an lenght check after each processed message garantees that > we do not underflow. The downside is that as soon an invalid length is > spotted the processing stops. Hmm, isn't NLMSG_OK already enough for this? It checks that length is = positive and above a certain threshold? > --- > monitor/nlmon.c | 5 +++++ > 1 file changed, 5 insertions(+) > = > diff --git a/monitor/nlmon.c b/monitor/nlmon.c > index 6b0afef8a497..5498e9fa0f23 100644 > --- a/monitor/nlmon.c > +++ b/monitor/nlmon.c > @@ -6941,6 +6941,11 @@ static bool nlmon_receive(struct l_io *io, void *u= ser_data) > nlmon_message(nlmon, tv, tp, nlmsg); > break; > } > + > + if (nlmsg->nlmsg_len & (NLMSG_ALIGNTO-1)) { > + printf("malformed packet\n"); > + break; > + } Also I'm not sure whether this really holds for messages that are = 'last'. There's no sense to check or enforce alignment in such a case/ > } > = > return true; > = Regards, -Denis --===============1920542019371304447==--