From: Dmitry Yakunin Date: Sat, 9 May 2020 16:52:01 +0000 (+0300) Subject: ss: add support for cgroup v2 information and filtering X-Git-Url: https://git.proxmox.com/?p=mirror_iproute2.git;a=commitdiff_plain;h=14f4bda590447179e2b338c6e494658d2f737e6e ss: add support for cgroup v2 information and filtering This patch introduces two new features: obtaining cgroup information and filtering sockets by cgroups. These features work based on cgroup v2 ID field in the socket (kernel should be compiled with CONFIG_SOCK_CGROUP_DATA). Cgroup information can be obtained by specifying --cgroup flag and now contains only pathname. For faster pathname lookups cgroup cache is implemented. This cache is filled on ss startup and missed entries are resolved and saved on the fly. Cgroup filter extends EXPRESSION and allows to specify cgroup pathname (relative or absolute) to obtain sockets attached only to this cgroup. Filter syntax: ss [ cgroup PATHNAME ] Examples: ss -a cgroup /sys/fs/cgroup/unified (or ss -a cgroup .) ss -a cgroup /sys/fs/cgroup/unified/cgroup1 (or ss -a cgroup cgroup1) v2: - style fixes (David Ahern) Signed-off-by: Dmitry Yakunin Signed-off-by: David Ahern --- diff --git a/man/man8/ss.8 b/man/man8/ss.8 index c80853f9..3b2559ff 100644 --- a/man/man8/ss.8 +++ b/man/man8/ss.8 @@ -286,6 +286,15 @@ Class id set by net_cls cgroup. If class is zero this shows priority set by SO_PRIORITY. .RE .TP +.B \-\-cgroup +Show cgroup information. Below fields may appear: +.RS +.P +.TP +.B cgroup +Cgroup v2 pathname. This pathname is relative to the mount point of the hierarchy. +.RE +.TP .B \-K, \-\-kill Attempts to forcibly close sockets. This option displays sockets that are successfully closed and silently skips sockets that the kernel does not support diff --git a/misc/ss.c b/misc/ss.c index ee840149..2a71317d 100644 --- a/misc/ss.c +++ b/misc/ss.c @@ -36,6 +36,7 @@ #include "namespace.h" #include "SNAPSHOT.h" #include "rt_names.h" +#include "cg_map.h" #include #include @@ -123,6 +124,7 @@ static int follow_events; static int sctp_ino; static int show_tipcinfo; static int show_tos; +static int show_cgroup; int oneline; enum col_id { @@ -798,6 +800,7 @@ struct sockstat { char *name; char *peer_name; __u32 mark; + __u64 cgroup_id; }; struct dctcpstat { @@ -1418,6 +1421,9 @@ static void sock_details_print(struct sockstat *s) if (s->mark) out(" fwmark:0x%x", s->mark); + + if (s->cgroup_id) + out(" cgroup:%s", cg_id_to_path(s->cgroup_id)); } static void sock_addr_print(const char *addr, char *delim, const char *port, @@ -1644,6 +1650,7 @@ struct aafilter { unsigned int iface; __u32 mark; __u32 mask; + __u64 cgroup_id; struct aafilter *next; }; @@ -1771,6 +1778,12 @@ static int run_ssfilter(struct ssfilter *f, struct sockstat *s) struct aafilter *a = (void *)f->pred; return (s->mark & a->mask) == a->mark; + } + case SSF_CGROUPCOND: + { + struct aafilter *a = (void *)f->pred; + + return s->cgroup_id == a->cgroup_id; } /* Yup. It is recursion. Sorry. */ case SSF_AND: @@ -1962,6 +1975,23 @@ static int ssfilter_bytecompile(struct ssfilter *f, char **bytecode) { a->mark, a->mask}, }; + return inslen; + } + case SSF_CGROUPCOND: + { + struct aafilter *a = (void *)f->pred; + struct instr { + struct inet_diag_bc_op op; + __u64 cgroup_id; + } __attribute__((packed)); + int inslen = sizeof(struct instr); + + if (!(*bytecode = malloc(inslen))) abort(); + ((struct instr *)*bytecode)[0] = (struct instr) { + { INET_DIAG_BC_CGROUP_COND, inslen, inslen + 4 }, + a->cgroup_id, + }; + return inslen; } default: @@ -2301,6 +2331,22 @@ void *parse_markmask(const char *markmask) return res; } +void *parse_cgroupcond(const char *path) +{ + struct aafilter *res; + __u64 id; + + id = get_cgroup2_id(path); + if (!id) + return NULL; + + res = malloc(sizeof(*res)); + if (res) + res->cgroup_id = id; + + return res; +} + static void proc_ctx_print(struct sockstat *s) { char *buf; @@ -3157,6 +3203,9 @@ static void parse_diag_msg(struct nlmsghdr *nlh, struct sockstat *s) s->mark = 0; if (tb[INET_DIAG_MARK]) s->mark = rta_getattr_u32(tb[INET_DIAG_MARK]); + s->cgroup_id = 0; + if (tb[INET_DIAG_CGROUP_ID]) + s->cgroup_id = rta_getattr_u64(tb[INET_DIAG_CGROUP_ID]); if (tb[INET_DIAG_PROTOCOL]) s->raw_prot = rta_getattr_u8(tb[INET_DIAG_PROTOCOL]); else @@ -3224,6 +3273,11 @@ static int inet_show_sock(struct nlmsghdr *nlh, out(" class_id:%#x", rta_getattr_u32(tb[INET_DIAG_CLASS_ID])); } + if (show_cgroup) { + if (tb[INET_DIAG_CGROUP_ID]) + out(" cgroup:%s", cg_id_to_path(rta_getattr_u64(tb[INET_DIAG_CGROUP_ID]))); + } + if (show_mem || (show_tcpinfo && s->type != IPPROTO_UDP)) { if (!oneline) out("\n\t"); @@ -5049,6 +5103,7 @@ static void _usage(FILE *dest) " --tipcinfo show internal tipc socket information\n" " -s, --summary show socket usage summary\n" " --tos show tos and priority information\n" +" --cgroup show cgroup information\n" " -b, --bpf show bpf filter socket information\n" " -E, --events continually display sockets as they are destroyed\n" " -Z, --context display process SELinux security contexts\n" @@ -5159,6 +5214,8 @@ static int scan_state(const char *state) /* Values of 'x' are already used so a non-character is used */ #define OPT_XDPSOCK 260 +#define OPT_CGROUP 261 + static const struct option long_opts[] = { { "numeric", 0, 0, 'n' }, { "resolve", 0, 0, 'r' }, @@ -5195,6 +5252,7 @@ static const struct option long_opts[] = { { "net", 1, 0, 'N' }, { "tipcinfo", 0, 0, OPT_TIPCINFO}, { "tos", 0, 0, OPT_TOS }, + { "cgroup", 0, 0, OPT_CGROUP }, { "kill", 0, 0, 'K' }, { "no-header", 0, 0, 'H' }, { "xdp", 0, 0, OPT_XDPSOCK}, @@ -5382,6 +5440,9 @@ int main(int argc, char *argv[]) case OPT_TOS: show_tos = 1; break; + case OPT_CGROUP: + show_cgroup = 1; + break; case 'K': current_filter.kill = 1; break; diff --git a/misc/ssfilter.h b/misc/ssfilter.h index f5b0bc8a..d85c084e 100644 --- a/misc/ssfilter.h +++ b/misc/ssfilter.h @@ -11,6 +11,7 @@ #define SSF_S_AUTO 9 #define SSF_DEVCOND 10 #define SSF_MARKMASK 11 +#define SSF_CGROUPCOND 12 #include @@ -25,3 +26,4 @@ int ssfilter_parse(struct ssfilter **f, int argc, char **argv, FILE *fp); void *parse_hostcond(char *addr, bool is_port); void *parse_devcond(char *name); void *parse_markmask(const char *markmask); +void *parse_cgroupcond(const char *path); diff --git a/misc/ssfilter.y b/misc/ssfilter.y index a901ae75..b4175795 100644 --- a/misc/ssfilter.y +++ b/misc/ssfilter.y @@ -36,7 +36,7 @@ static void yyerror(char *s) %} -%token HOSTCOND DCOND SCOND DPORT SPORT LEQ GEQ NEQ AUTOBOUND DEVCOND DEVNAME MARKMASK FWMARK +%token HOSTCOND DCOND SCOND DPORT SPORT LEQ GEQ NEQ AUTOBOUND DEVCOND DEVNAME MARKMASK FWMARK CGROUPCOND CGROUPPATH %left '|' %left '&' %nonassoc '!' @@ -156,6 +156,14 @@ expr: '(' exprlist ')' { $$ = alloc_node(SSF_NOT, alloc_node(SSF_MARKMASK, $3)); } + | CGROUPPATH eq CGROUPCOND + { + $$ = alloc_node(SSF_CGROUPCOND, $3); + } + | CGROUPPATH NEQ CGROUPCOND + { + $$ = alloc_node(SSF_NOT, alloc_node(SSF_CGROUPCOND, $3)); + } | AUTOBOUND { $$ = alloc_node(SSF_S_AUTO, NULL); @@ -276,6 +284,10 @@ int yylex(void) tok_type = FWMARK; return FWMARK; } + if (strcmp(curtok, "cgroup") == 0) { + tok_type = CGROUPPATH; + return CGROUPPATH; + } if (strcmp(curtok, ">=") == 0 || strcmp(curtok, "ge") == 0 || strcmp(curtok, "geq") == 0) @@ -318,6 +330,14 @@ int yylex(void) } return MARKMASK; } + if (tok_type == CGROUPPATH) { + yylval = (void*)parse_cgroupcond(curtok); + if (yylval == NULL) { + fprintf(stderr, "Cannot parse cgroup %s.\n", curtok); + exit(1); + } + return CGROUPCOND; + } yylval = (void*)parse_hostcond(curtok, tok_type == SPORT || tok_type == DPORT); if (yylval == NULL) { fprintf(stderr, "Cannot parse dst/src address.\n");