aboutsummaryrefslogtreecommitdiff
path: root/src
AgeCommit message (Collapse)AuthorFilesLines
2020-12-01socket: consume empty packetsGiuseppe Scrivano1-3/+0
it fixes a tightloop when a packet with len==0 is received. Closes: https://github.com/rootless-containers/slirp4netns/issues/227 Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
2020-08-27Add G_GNUC_PRINTF to local function slirp_vsnprintfStefan Weil1-0/+1
Signed-off-by: Stefan Weil <sw@weilnetz.de>
2020-08-19sosendoob: better document what urgc is used forSamuel Thibault1-3/+4
2020-08-18TCPIPHDR_DELTA: Fix potential negative valueSamuel Thibault1-2/+2
sizeof() returns a size_t so the tcpiphdr / ip+tcphdr difference will be a size_t and always be >= 0, while this intended to detect the difference getting < 0. This is actually a no-op with the current code because it currently has tcpiphdr bigger than ip+tcphdr. Spotted by Coverity: CID 212435. Spotted by Coverity: CID 212440. Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-07-17ip_stripoptions use memmoveDr. David Alan Gilbert1-1/+1
ip_stripoptions is moving data long in the same buffer; that's undefined with memcpy, use memmove. Buglink: https://bugs.launchpad.net/qemu/+bug/1878043 Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2020-07-08util: do not silently truncateMarc-André Lureau1-1/+1
snprintf() always nul-terminate. The return value is the number of business bytes that would be produced if the buffer was large enough. If it returns N for a N size buffer, it means truncation occurred (and we lost one business byte). Related to: #22 Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-07-03Fix MTU checkRalf Haferkamp1-1/+1
The size for Header has to be accounted for as well.
2020-07-03Drop bogus IPv6 messagesRalf Haferkamp1-0/+7
Drop IPv6 message shorter than what's mentioned in the payload length header (+ the size of the IPv6 header). They're invalid an could lead to data leakage in icmp6_send_echoreply().
2020-06-01Remove unnecessary breakPhilippe Mathieu-Daudé3-5/+0
The code is unreachable, so no need to break. This silence static analyzer warnings. Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-06-01Fix constness warningsPhilippe Mathieu-Daudé5-5/+9
Fix the following GCC warnings: src/ncsi.c: In function ‘ncsi_input’: src/ncsi.c:139:31: error: cast discards ‘const’ qualifier from pointer target type [-Werror=cast-qual] 139 | struct ncsi_pkt_hdr *nh = (struct ncsi_pkt_hdr *)(pkt + ETH_HLEN); | ^ src/dnssearch.c: In function ‘translate_dnssearch’: src/dnssearch.c:242:33: error: cast discards ‘const’ qualifier from pointer target type [-Werror=cast-qual] 242 | num_domains = g_strv_length((GStrv)names); | ^ src/slirp.c: In function ‘arp_input’: src/slirp.c:747:31: error: cast discards ‘const’ qualifier from pointer target type [-Werror=cast-qual] 747 | struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN); | ^ src/dnssearch.c: In function ‘translate_dnssearch’: src/dnssearch.c:242:33: error: cast discards ‘const’ qualifier from pointer target type [-Werror=cast-qual] 242 | num_domains = g_strv_length((const GStrv)names); | ^ src/slirp.c: In function ‘arp_input’: src/slirp.c:764:48: error: passing argument 3 of ‘arp_table_add’ discards ‘const’ qualifier from pointer target type [-Werror=discarded-qualifiers] 764 | arp_table_add(slirp, ah->ar_sip, ah->ar_sha); | ~~^~~~~~~~ In file included from src/slirp.c:25: src/slirp.h:101:60: note: expected ‘uint8_t *’ {aka ‘unsigned char *’} but argument is of type ‘const unsigned char *’ 101 | void arp_table_add(Slirp *slirp, uint32_t ip_addr, uint8_t ethaddr[ETH_ALEN]); | ~~~~~~~~^~~~~~~~~~~~~~~~~ src/slirp.c:783:48: error: passing argument 3 of ‘arp_table_add’ discards ‘const’ qualifier from pointer target type [-Werror=discarded-qualifiers] 783 | arp_table_add(slirp, ah->ar_sip, ah->ar_sha); | ~~^~~~~~~~ In file included from src/slirp.c:25: src/slirp.h:101:60: note: expected ‘uint8_t *’ {aka ‘unsigned char *’} but argument is of type ‘const unsigned char *’ 101 | void arp_table_add(Slirp *slirp, uint32_t ip_addr, uint8_t ethaddr[ETH_ALEN]); | ~~~~~~~~^~~~~~~~~~~~~~~~~ src/slirp.c:804:44: error: passing argument 3 of ‘arp_table_add’ discards ‘const’ qualifier from pointer target type [-Werror=discarded-qualifiers] 804 | arp_table_add(slirp, ah->ar_sip, ah->ar_sha); | ~~^~~~~~~~ In file included from src/slirp.c:25: src/slirp.h:101:60: note: expected ‘uint8_t *’ {aka ‘unsigned char *’} but argument is of type ‘const unsigned char *’ 101 | void arp_table_add(Slirp *slirp, uint32_t ip_addr, uint8_t ethaddr[ETH_ALEN]); | ~~~~~~~~^~~~~~~~~~~~~~~~~ Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-06-01Fix win32 builds by using the SLIRP_PACKED definitionPhilippe Mathieu-Daudé1-1/+3
A packed struct needs different gcc attributes for compilations with MinGW compilers because glib-2.0 adds compiler flag -mms-bitfields which modifies the packing algorithm. Attribute gcc_struct reverses the negative effects of -mms-bitfields. We already have the SLIRP_PACKED definition for that, use it. Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-05-27Check lseek() for failureJindrich Novy1-1/+3
Error: CHECKED_RETURN (CWE-252): [#def26] libslirp-4.3.0/src/tftp.c:121: check_return: Calling "lseek(spt->fd, block_nr * spt->block_size, 0)" without checking return value. This library function may fail and return an error code. 119| 120| if (len) { 121|-> lseek(spt->fd, block_nr * spt->block_size, SEEK_SET); 122| 123| bytes_read = read(spt->fd, buf, len); Signed-off-by: Jindrich Novy <jnovy@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-05-27Be sure to initialize sockaddr structureJindrich Novy1-0/+1
Error: UNINIT (CWE-457): [#def30] libslirp-4.3.0/src/udp.c:325: var_decl: Declaring variable "addr" without initializer. libslirp-4.3.0/src/udp.c:342: uninit_use_in_call: Using uninitialized value "addr". Field "addr.sin_zero" is uninitialized when calling "bind". Signed-off-by: Jindrich Novy <jnovy@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-05-27Use secure string copy to avoid overflowJindrich Novy1-2/+2
Error: STRING_OVERFLOW (CWE-120): [#def2] libslirp-4.3.0/src/ip_icmp.c:277: fixed_size_dest: You might overrun the 20-character fixed-size string "bufa" by copying the return value of "inet_ntoa" without checking the length. 275| if (slirp_debug & DBG_MISC) { 276| char bufa[20], bufb[20]; 277|-> strcpy(bufa, inet_ntoa(ip->ip_src)); 278| strcpy(bufb, inet_ntoa(ip->ip_dst)); 279| DEBUG_MISC(" %.16s to %.16s", bufa, bufb); Error: STRING_OVERFLOW (CWE-120): [#def3] libslirp-4.3.0/src/ip_icmp.c:278: fixed_size_dest: You might overrun the 20-character fixed-size string "bufb" by copying the return value of "inet_ntoa" without checking the length. 276| char bufa[20], bufb[20]; 277| strcpy(bufa, inet_ntoa(ip->ip_src)); 278|-> strcpy(bufb, inet_ntoa(ip->ip_dst)); 279| DEBUG_MISC(" %.16s to %.16s", bufa, bufb); 280| } Signed-off-by: Jindrich Novy <jnovy@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-05-27Fix possible infinite loops and use-after-freeJindrich Novy2-3/+9
Error: USE_AFTER_FREE (CWE-416): [#def1] libslirp-4.3.0/src/ip_icmp.c:79: freed_arg: "icmp_detach" frees "slirp->icmp.so_next". libslirp-4.3.0/src/ip_icmp.c:79: deref_arg: Calling "icmp_detach" dereferences freed pointer "slirp->icmp.so_next". 77| { 78| while (slirp->icmp.so_next != &slirp->icmp) { 79|-> icmp_detach(slirp->icmp.so_next); 80| } 81| } Error: USE_AFTER_FREE (CWE-416): [#def27] libslirp-4.3.0/src/udp.c:56: freed_arg: "udp_detach" frees "slirp->udb.so_next". libslirp-4.3.0/src/udp.c:56: deref_arg: Calling "udp_detach" dereferences freed pointer "slirp->udb.so_next". 54| { 55| while (slirp->udb.so_next != &slirp->udb) { 56|-> udp_detach(slirp->udb.so_next); 57| } 58| } Signed-off-by: Jindrich Novy <jnovy@redhat.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-04-06Fix use-afte-free in ip_reass() (CVE-2020-1983)Marc-André Lureau1-4/+2
The q pointer is updated when the mbuf data is moved from m_dat to m_ext. m_ext buffer may also be realloc()'ed and moved during m_cat(): q should also be updated in this case. Reported-by: Aviv Sasson <asasson@paloaltonetworks.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-03-25libslirp.h: fix SlirpConfig v3 documentationAkihiro Suda1-0/+3
Follow-up to 1021b0dc38d39f1dc95a296fe3e05a24a087cdc6 (https://gitlab.freedesktop.org/slirp/libslirp/-/merge_requests/31) Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
2020-03-24limit vnameserver_addr to port 535eraph1-2/+2
Fixes #16 Signed-off-by: 5eraph <5eraph@protonmail.com>
2020-03-23disable_dns option5eraph4-3/+12
Fixes #16 Signed-off-by: 5eraph <5eraph@protonmail.com>
2020-03-20build-sys: fix NetBSD build regressionMarc-André Lureau1-0/+5
Fixes: 09d410adbff5422b7ba7596bce0ca71f9f807ea9 ("allow custom MTU") Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-17Merge branch 'vcs-version' into 'master'Marc-André Lureau2-4/+2
Teach slirp_version_string() to return vcs version See merge request slirp/libslirp!34
2020-03-17Teach slirp_version_string() to return vcs versionMarc-André Lureau2-4/+2
Meson build will use a vcs-generate version, while Makefile will always use -git version, since it is only intended for submodule usage. Eventually can be improved if needed. Fixes: https://gitlab.freedesktop.org/slirp/libslirp/issues/17 Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-17Revert "socket: remove need for extra scope_id variable"Marc-André Lureau1-1/+6
Oops, it turns out the variable is there for portability reasons. This reverts commit d65f3030a82743bf506b0611a6a1a0358ea5d52b. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-16socket: do not fallback on loopback addr for addresses in our mask/prefixMarc-André Lureau1-16/+12
Currently, any address within the subnetwork will fallback on loopback. It seems it has always been like that, but it seems wrong, and I don't see a good reason to keep it this way. Fortunately, lack of ARP reply made this unusable in practice, so we shouldn't break much existing users. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-16socket: do not fallback on host loopback if get_dns_addr() failedMarc-André Lureau1-6/+2
Somewhat related to #16, but not as restrictive. (imho, it should be possible to access any port on the given DNS IP, not just 53) Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-16socket: remove need for extra scope_id variableMarc-André Lureau1-3/+1
The value is only set on success. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-16socket: factor out sotranslate ipv4/ipv6 handlingMarc-André Lureau1-42/+54
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-16socket: remove extra label and variableMarc-André Lureau1-9/+6
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-03-02Use specific outbound IP address5eraph8-1/+58
Fixes #14 Signed-off-by: 5eraph <bcervenka@protonmail.com>
2020-02-03Merge branch 'slirp-fmt' into 'master'Marc-André Lureau3-7/+9
misc: slirp_fmt*() improvements See merge request slirp/libslirp!28
2020-01-30state: fix loading of guestfwd stateAnders Waldenborg1-0/+2
The refactoring done in commit d181d14b "slirp: use a dedicated field for chardev pointer" forgot to change one place in slirp_state_load where 'ex_exec' was used to store the chardev ptr. This broke loading of saved state. Later commit 4f38cfb5 "slirp: remove unused EMU_RSH" removed this line all together, as it now looked like it didn't do anything. This commit ensures that guestfwd is properly setup on the socket when loading state. Signed-off-by: Anders Waldenborg <anders@0x63.nu>
2020-01-27util: add gnuc format function attribute to slirp_fmt*Marc-André Lureau1-2/+4
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
2020-01-27Use g_snprintf()Marc-André Lureau2-3/+3
The GLib impl guarantees GNU compatible format strings, which fixes the horror of Windows platform format strings. Suggested-by: Daniel P. Berrangé <berrange@redhat.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-01-27misc: improve error reportMarc-André Lureau1-2/+2
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-01-27tcp_emu: fix unsafe snprintf() usagesMarc-André Lureau1-23/+21
Various calls to snprintf() assume that snprintf() returns "only" the number of bytes written (excluding terminating NUL). https://pubs.opengroup.org/onlinepubs/9699919799/functions/snprintf.html#tag_16_159_04 "Upon successful completion, the snprintf() function shall return the number of bytes that would be written to s had n been sufficiently large excluding the terminating null byte." Before patch ce131029, if there isn't enough room in "m_data" for the "DCC ..." message, we overflow "m_data". After the patch, if there isn't enough room for the same, we don't overflow "m_data", but we set "m_len" out-of-bounds. The next time an access is bounded by "m_len", we'll have a buffer overflow then. Use slirp_fmt*() to fix potential OOB memory access. Reported-by: Laszlo Ersek <lersek@redhat.com> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-7-marcandre.lureau@redhat.com>
2020-01-27tcp_ctl: use slirp_fmt()Marc-André Lureau1-3/+2
Make it safer to OOB (sb_cc must not go out of sb_data), warn on truncation, abort on error. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-6-marcandre.lureau@redhat.com>
2020-01-27tftp: use slirp_fmt0()Marc-André Lureau1-6/+2
Make it OOB-safe, warn on truncation, always \0-end, abort on error. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-5-marcandre.lureau@redhat.com>
2020-01-27misc: use slirp_fmt0()Marc-André Lureau1-6/+6
Those are safe and should never fail. Nevertheless, use slirp_snfillf0() for more safety. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-4-marcandre.lureau@redhat.com>
2020-01-27dhcpv6: use slirp_fmt()Marc-André Lureau1-7/+6
Warn if result is truncated, return bytes actually written (excluding \0). Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-3-marcandre.lureau@redhat.com>
2020-01-27util: add slirp_fmt() helpersMarc-André Lureau2-0/+65
Various calls to snprintf() in libslirp assume that snprintf() returns "only" the number of bytes written (excluding terminating NUL). https://pubs.opengroup.org/onlinepubs/9699919799/functions/snprintf.html#tag_16_159_04 "Upon successful completion, the snprintf() function shall return the number of bytes that would be written to s had n been sufficiently large excluding the terminating null byte." Introduce slirp_fmt() that handles several pathological cases the way libslirp usually expect: - treat error as fatal (instead of silently returning -1) - fmt0() will always \0 end - return the number of bytes actually written (instead of what would have been written, which would usually result in OOB later), including the ending \0 for fmt0() - warn if truncation happened (instead of ignoring) Other less common cases can still be handled with strcpy/snprintf() etc. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200127092414.169796-2-marcandre.lureau@redhat.com>
2020-01-22tcp_emu: add more fixme/warnings commentsMarc-André Lureau1-0/+3
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-01-22tftp: clarify what is actually OACK m_lenMarc-André Lureau1-2/+1
The current computation is a bit convoluted, and doesn't reflect >0. What is actually computed is sizeof(): struct tftp_t { struct udphdr udp; uint16_t tp_op; union { ... char tp_buf[TFTP_BLOCKSIZE_MAX + 2]; } x; } - sizeof(struct udphdr) == udp field - (TFTP_BLOCKSIZE_MAX + 2) == tp_buf field + n What remains is: G_SIZEOF_MEMBER(struct tftp_t, tp_op) + n. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-01-22bootp: replace simple snprintf() with strcpy()Marc-André Lureau1-3/+4
Minor code simplification. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
2020-01-22bootp: remove extra castMarc-André Lureau2-2/+2
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-01-22Check bootp_filename is not going to be truncatedMarc-André Lureau1-0/+3
If the given bootp_filename is too long, it is silently truncated in bootp.c snprintf(). Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-01-22util: add G_SIZEOF_MEMBER() macroMarc-André Lureau1-0/+4
I am (overly?) optimistic this macro will be added to glib: https://gitlab.gnome.org/GNOME/glib/merge_requests/1333 Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
2020-01-14slirp: tftp: restrict relative path accessPrasad J Pandit1-2/+7
tftp restricts relative or directory path access on Linux systems. Apply same restrictions on Windows systems too. It helps to avoid directory traversal issue. Fixes: https://bugs.launchpad.net/qemu/+bug/1812451 Reported-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200113121431.156708-1-ppandit@redhat.com>
2020-01-12slirp: use correct size while emulating commandsPrasad J Pandit1-5/+4
While emulating services in tcp_emu(), it uses 'mbuf' size 'm->m_size' to write commands via snprintf(3). Use M_FREEROOM(m) size to avoid possible OOB access. Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200109094228.79764-3-ppandit@redhat.com>
2020-01-12slirp: use correct size while emulating IRC commandsPrasad J Pandit1-5/+6
While emulating IRC DCC commands, tcp_emu() uses 'mbuf' size 'm->m_size' to write DCC commands via snprintf(3). This may lead to OOB write access, because 'bptr' points somewhere in the middle of 'mbuf' buffer, not at the start. Use M_FREEROOM(m) size to avoid OOB access. Reported-by: Vishnu Dev TJ <vishnudevtj@gmail.com> Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org> Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org> Message-Id: <20200109094228.79764-2-ppandit@redhat.com>
2020-01-09tcp_emu: Fix oob accessSamuel Thibault1-0/+7
The main loop only checks for one available byte, while we sometimes need two bytes.
generated by cgit v1.1 at 2024-07-24 11:18:33 +0000