Commits on Source (47)
-
Marc-André Lureau authored113a219a
-
Samuel Thibault authored
When the first fragment does not fit in the preallocated buffer, q will already be pointing to the ext buffer, so we mustn't try to update it. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
126c04ac -
Akihiro Suda authored
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
43b13d5b -
Akihiro Suda authored
From https://github.com/rootless-containers/slirp4netns/blob/4889f5299f407d7d7566c76a3b8b5f71c99b6db5/qemu_patches/0002-slirp-allow-custom-MTU.patch Original commits: * https://github.com/rootless-containers/slirp4netns/commit/ea630a7e945cf538184ff1b1b4bd7b8ddc01993e * https://github.com/rootless-containers/slirp4netns/commit/1508a66c93c223555f08651592dde3d2d708b166 * https://github.com/rootless-containers/slirp4netns/commit/19f3f41df4066d6103e6f882500e24db7ea7d9e1 * https://github.com/rootless-containers/slirp4netns/commit/a11abedafcc627ef0657999e63b211b0f26d4c02 * https://github.com/rootless-containers/slirp4netns/commit/2adbd7c449944d3b837164c86eedd3dcabbba1a6 Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
09d410ad -
Akihiro Suda authored
From https://github.com/rootless-containers/slirp4netns/blob/4889f5299f407d7d7566c76a3b8b5f71c99b6db5/qemu_patches/0003-slirp-add-disable_host_loopback-prohibit-connections.patch Original commits: * https://github.com/rootless-containers/slirp4netns/commit/6325473781bb344c225f54e2d28800fb0619d7ee * https://github.com/rootless-containers/slirp4netns/commit/13b24026867d4c30d5d1465ac82e3bb890bf4caa Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
95932a68 -
Akihiro Suda authored
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
09ea219f -
Marc-André Lureau authored30804efc
-
Samuel Thibault authored
Closes #10 Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
e0be8043 -
Samuel Thibault authored
ip_reass: explain why we should not always update the q pointer Closes #10 See merge request !9
ea91ceef -
Akihiro Suda authored
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
5367fa15 -
Akihiro Suda authored
tcp_emu() is known to have caused several CVEs, and not useful today in most cases. https://nvd.nist.gov/vuln/detail/CVE-2019-6778 https://nvd.nist.gov/vuln/detail/CVE-2019-9824 The feature can be still enabled by setting SlirpConfig.enable_emu to true. Closes #11 Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
07c2a44b -
Akihiro Suda authored
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
db32b649 -
Samuel Thibault authored
emu: disable by default Closes #11 See merge request !10
6b6d4f76 -
Samuel Thibault authored
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
3ff51a35 -
Marc-André Lureau authored
fix a typo in a comment See merge request !11
76462e2f -
Samuel Thibault authored
Using ip_deq after m_free might read pointers from an allocation reuse. This would be difficult to exploit, but that is still related with CVE-2019-14378 which generates fragmented IP packets that would trigger this issue and at least produce a DoS. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
c5927943 -
Samuel Thibault authored
ip_reass: Fix use after free See merge request !12
d203c81b -
Marc-André Lureau authored
Using library() allows to build either static or shared library. Declare a dependency, so parent projects can link with it easily: slirp_proj = subproject('slirp') slirp = slirp_proj.get_variable('libslirp_dep') Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
49729df7 -
Marc-André Lureau authored
meson: make it subproject friendly See merge request !14
ad36c04a -
Samuel Thibault authored
This is needed for using an external DHCP server Fixes #9
4657a267 -
Samuel Thibault authored
That can show up with DHCP packets. Fixes #9 Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
db65c3cb -
g_shell_parse_argv does only tokenization, and no replacement, so it is safe to use it here. This quesion arised when modifying QEMU because the new version 4 of Samba disables version 1 of the SMB protocols, to run old Win clients I am developing a patch that gets the value of the environment variable SMBDOPTIONS and appends it to the smbd command line; it allows the user to specify additional samba daemon parameters before starting qemu. Example: export SMBDOPTIONS="--option='server min protocol=CORE' -d 4" Signed-off-by: Jordi Pujol Palomer <jordipujolp@gmail.com>
76031015 -
Marc-André Lureau authored
Introduced in previous commit: ../src/misc.c: In function ‘fork_exec’: ../src/misc.c:184:9: warning: assignment to ‘GError *’ {aka ‘struct _GError *’} from ‘gboolean’ {aka ‘int’} makes pointer from integer without a cast [-Wint-conversion] 184 | err = g_shell_parse_argv(ex, &argc, &argv, &err); | ^ ../src/misc.c:173:14: warning: unused variable ‘ret’ [-Wunused-variable] 173 | gboolean ret; | ^~~ Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
48663b31 -
Samuel Thibault authored
misc: fix compilation warnings See merge request !18
fe80a346 -
Giuseppe Scrivano authored
avoid using getpeername(2) if the socket was already closed for writing, as it will report the socket as disconnected. Using getsockopt instead ensures there is no error returned. Closes: #12 Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
-
Marc-André Lureau authored
socket: avoid getpeername after shutdown(SHUT_WR) Closes #12 See merge request !19
1cfbe0fe -
Jindrich Novy authored
Signed-off-by: Jindrich Novy <jnovy@redhat.com> [ Marc-André - modified to use a temporary variable ] Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
d171af37 -
Marc-André Lureau authored
The only sbcopy() caller is tcp_output(). There, len is constrained to be 0 <= len <= sb_cc. Let's add some assert to avoid potential undefined behaviour (the function didn't return the actual number of bytes copied). Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
d0e97384 -
Marc-André Lureau authored
We shouldn't be reading undefined data, check that the data to read remains within sb_cc limit. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
db1fabb7 -
Marc-André Lureau authored
Let see if it happens, and drop it eventually some day. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
61f54d2e -
Marc-André Lureau authored
glib mem functions are already used in various places. Let's not mix the two, and instead abort on OOM conditions. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
3a494648 -
Marc-André Lureau authored
Now that tcp_newtcpcb() always returns != NULL. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
ab9a34aa -
Marc-André Lureau authored
sbreserve() will always succeed or abort(). Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
a4412bae -
Samuel Thibault authored
arp: Allow 0.0.0.0 destination address Closes #9 See merge request !16
0d337b53 -
Marc-André Lureau authored
Negative values wouldn't make sense in those functions and could lead to weird results. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
8716c0e4 -
Marc-André Lureau authored
realloc/g_realloc() allocates memory if given ptr is NULL. Note: This changes a bit the code, since now sb_cc is always reset to 0, even if old and new value are the same. This seems more coherent, but may have weird side-effects if code relies on it. Reviewing usage of sbreserve() reveals that it is used before the socket buffer receives any data, at tcp_input() socket creation time, and during tcp_mss() which is earlier in TCP socket state. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
be26d1cc -
Marc-André Lureau authored
This has also the side-effect of silencing a false-positive in scan-build. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
d01e32c4 -
Marc-André Lureau authored
Silence: src/vmstate.c:324:17: warning: Value stored to 'ret' is never read Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
cf45f131 -
Marc-André Lureau authored
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
d7846125 -
Samuel Thibault authored
Make host receive broadcast packets Closes #9 See merge request !15
ccfc7f25 -
PanNengyuan authored
qemu crashes with a segfault (NULL pointer access in tcp_sockclosed), tp = tcp_close(tp) will free tp and set tp to NULL, then tcp_output(tp) access the null pointer(tp). This fixes: 384 break; 385 } CID 68914397: (NULL_RETURNS) 386. dereference: Dereferencing a pointer that might be "NULL" "tp" when calling "tcp_output". 386 tcp_output(tp); 387} Reported-by: Euler Robot <euler.robot@huawei.com> Signed-off-by: PanNengyuan <pannengyuan@huawei.com> Message-Id: <1574644852-24440-1-git-send-email-pannengyuan@huawei.com> Fixes: 804f441a Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
e03380e5 -
Marc-André Lureau authored
Mem cleanups See merge request !20
67a6170c -
Marc-André Lureau authored
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
9dbc96fd -
Marc-André Lureau authored
libtool versions must be translated to maj.min.rev, where maj = current - age. Also fix lt_current usage, quoting Samuel Thibault: soversion is not the same as lt_current. soversion must be lt_current - lt_age, so that the soname stays the same when binary compatibility is preserved (here, 0), and gets bumped on compatibility break. Yes, on compatibility break the bump will be big. That's because non-soname OSes the compatibility support is different. The details are hairy, yes :) Instead, rely on meson doing the right thing with soversion by default. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
339f791f -
Marc-André Lureau authored
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
bfa16aba -
Marc-André Lureau authored
Release v4.1.0 See merge request !21
CHANGELOG.md
0 → 100644