From 7a27810aecce6e014100dbbc3c7961d769082c1e Mon Sep 17 00:00:00 2001 From: vmallika Date: Mon, 24 Aug 2015 08:07:14 +0530 Subject: cli: on error invoke cli_cmd_broadcast_response function in separate thread There is a problem in current CLI framework CLI holds the lock when processing command. When processing quota list command, below sequence of steps executed in the same thread and causing deadlock 1) CLI holds the lock 2) Send rpc_clnt_submit request to quotad for quota usage 3) If quotad is down, rpc_clnt_submit invokes cbk function with error 4) cbk function cli_quotad_getlimit_cbk tries to hold lock to broadcast the results and hangs, because same thread has already holding the lock This patch fixes the problem by creating seperate thread for broadcasting the result Change-Id: I53be006eadf6aaf348083d9168535530d70a8ab3 BUG: 1242819 Signed-off-by: vmallika Reviewed-on: http://review.gluster.org/11990 Tested-by: Gluster Build System Tested-by: NetBSD Build System Reviewed-by: Krishnan Parthasarathi --- cli/src/cli-cmd-volume.c | 12 ------- cli/src/cli-rpc-ops.c | 94 ++++++++++++++++++++++++++++++++++++++++++++---- cli/src/cli.c | 1 + cli/src/cli.h | 1 + 4 files changed, 89 insertions(+), 19 deletions(-) (limited to 'cli/src') diff --git a/cli/src/cli-cmd-volume.c b/cli/src/cli-cmd-volume.c index e26f5435c63..040e255f89d 100644 --- a/cli/src/cli-cmd-volume.c +++ b/cli/src/cli-cmd-volume.c @@ -1376,18 +1376,6 @@ cli_cmd_quota_handle_list_all (const char **words, dict_t *options) CLI_LOCAL_INIT (local, words, frame, xdata); proc = &cli_quotad_clnt.proctable[GF_AGGREGATOR_GETLIMIT]; - if (!(global_state->mode & GLUSTER_MODE_XML)) { - print_quota_list_header (type); - } else { - ret = cli_xml_output_vol_quota_limit_list_begin - (local, 0, 0, NULL); - if (ret) { - gf_log ("cli", GF_LOG_ERROR, "Error in printing " - "xml output"); - goto out; - } - } - gfid_str = GF_CALLOC (1, gf_common_mt_char, 64); if (!gfid_str) { ret = -1; diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c index c10554daaee..3df9226d9e9 100644 --- a/cli/src/cli-rpc-ops.c +++ b/cli/src/cli-rpc-ops.c @@ -3232,7 +3232,8 @@ out: } int -print_quota_list_from_quotad (call_frame_t *frame, dict_t *rsp_dict) +print_quota_list_from_quotad (call_frame_t *frame, dict_t *rsp_dict, + int32_t list_count) { char *path = NULL; char *default_sl = NULL; @@ -3244,11 +3245,11 @@ print_quota_list_from_quotad (call_frame_t *frame, dict_t *rsp_dict) quota_limits_t *size_limits = NULL; int32_t type = 0; + GF_ASSERT (frame); + local = frame->local; gd_rsp_dict = local->dict; - GF_ASSERT (frame); - ret = dict_get_int32 (rsp_dict, "type", &type); if (ret) { gf_log ("cli", GF_LOG_ERROR, "Failed to get type"); @@ -3306,12 +3307,37 @@ print_quota_list_from_quotad (call_frame_t *frame, dict_t *rsp_dict) goto out; } + if (list_count == 0) { + if (!(global_state->mode & GLUSTER_MODE_XML)) { + print_quota_list_header (type); + } else { + ret = cli_xml_output_vol_quota_limit_list_begin + (local, 0, 0, NULL); + if (ret) { + gf_log ("cli", GF_LOG_ERROR, "Error in " + "printing xml output"); + goto out; + } + } + } + ret = print_quota_list_output (local, path, default_sl, &limits, &used_space, type); out: return ret; } +void* +cli_cmd_broadcast_response_detached (void *opaque) +{ + int32_t ret = 0; + + ret = (intptr_t) opaque; + cli_cmd_broadcast_response (ret); + + return NULL; +} + int cli_quotad_getlimit_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) @@ -3321,12 +3347,41 @@ cli_quotad_getlimit_cbk (struct rpc_req *req, struct iovec *iov, int ret = -1; dict_t *dict = NULL; call_frame_t *frame = NULL; + cli_local_t *local = NULL; + dict_t *gd_rsp_dict = NULL; + int32_t list_count = 0; + pthread_t th_id = {0, }; - if (-1 == req->rpc_status) { + frame = myframe; + GF_ASSERT (frame); + + local = frame->local; + gd_rsp_dict = local->dict; + + LOCK (&local->lock); + { + ret = dict_get_int32 (gd_rsp_dict, "quota-list-count", + &list_count); + if (ret) + list_count = 0; + ret = dict_set_int32 (gd_rsp_dict, "quota-list-count", + list_count + 1); + } + UNLOCK (&local->lock); + + if (ret) { + gf_log ("cli", GF_LOG_ERROR, "Failed to set " + "quota-list-count in dict"); goto out; } - frame = myframe; + if (-1 == req->rpc_status) { + if (list_count == 0) + cli_err ("Connection failed. Please check if quota " + "daemon is operational."); + ret = -1; + goto out; + } ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_cli_rsp); if (ret < 0) { @@ -3357,11 +3412,36 @@ cli_quotad_getlimit_cbk (struct rpc_req *req, struct iovec *iov, "unserialize req-buffer to dictionary"); goto out; } - print_quota_list_from_quotad (frame, dict); + print_quota_list_from_quotad (frame, dict, list_count); } out: - cli_cmd_broadcast_response (ret); + /* Bad Fix: CLI holds the lock to process a command. + * When processing quota list command, below sequence of steps executed + * in the same thread and causing deadlock + * + * 1) CLI holds the lock + * 2) Send rpc_clnt_submit request to quotad for quota usage + * 3) If quotad is down, rpc_clnt_submit invokes cbk function with error + * 4) cbk function cli_quotad_getlimit_cbk invokes + * cli_cmd_broadcast_response which tries to hold lock to broadcast + * the results and hangs, because same thread has already holding + * the lock + * + * Broadcasting response in a seperate thread which is not a + * good fix. This needs to be re-visted with better solution + */ + if (ret == -1) { + ret = pthread_create (&th_id, NULL, + cli_cmd_broadcast_response_detached, + (void *)-1); + if (ret) + gf_log ("cli", GF_LOG_ERROR, "pthread_create failed: " + "%s", strerror (errno)); + } else { + cli_cmd_broadcast_response (ret); + } + if (dict) dict_unref (dict); diff --git a/cli/src/cli.c b/cli/src/cli.c index 3a27a3e7085..5707d3fbdd7 100644 --- a/cli/src/cli.c +++ b/cli/src/cli.c @@ -652,6 +652,7 @@ cli_local_get () cli_local_t *local = NULL; local = GF_CALLOC (1, sizeof (*local), cli_mt_cli_local_t); + LOCK_INIT (&local->lock); return local; } diff --git a/cli/src/cli.h b/cli/src/cli.h index 965ce3f9ee4..deadd2d0a5d 100644 --- a/cli/src/cli.h +++ b/cli/src/cli.h @@ -146,6 +146,7 @@ struct cli_local { xmlDocPtr doc; int vol_count; #endif + gf_lock_t lock; }; struct cli_volume_status { -- cgit