From: Olga Kornievskaia <[email protected]>
Add support for handling upcalls on the new "nfsd4_cb" directory pipes.
Only new kernels (2.6.29) have support for this new pipe directory.
(The need for this new pipe directory will go away with NFSv4.1 where
the callback can be done on the same connection as the fore-channel.)
Signed-off-by: Olga Kornievskaia <[email protected]>
Signed-off-by: Kevin Coffman <[email protected]>
---
utils/gssd/gssd.c | 5 +++
utils/gssd/gssd.h | 2 +
utils/gssd/gssd_main_loop.c | 18 ++++++++++-
utils/gssd/gssd_proc.c | 73 +++++++++++++++++++++++++++++--------------
4 files changed, 74 insertions(+), 24 deletions(-)
diff --git a/utils/gssd/gssd.c b/utils/gssd/gssd.c
index f6949db..7a23362 100644
--- a/utils/gssd/gssd.c
+++ b/utils/gssd/gssd.c
@@ -57,6 +57,7 @@
char pipefs_dir[PATH_MAX] = GSSD_PIPEFS_DIR;
char pipefs_nfsdir[PATH_MAX] = GSSD_PIPEFS_DIR;
+char pipefs_nfscbdir[PATH_MAX] = GSSD_PIPEFS_DIR;
char keytabfile[PATH_MAX] = GSSD_DEFAULT_KEYTAB_FILE;
char ccachedir[PATH_MAX] = GSSD_DEFAULT_CRED_DIR;
char *ccachesearch[GSSD_MAX_CCACHE_SEARCH + 1];
@@ -163,6 +164,10 @@ main(int argc, char *argv[])
pipefs_dir, GSSD_SERVICE_NAME);
if (pipefs_nfsdir[sizeof(pipefs_nfsdir)-1] != '\0')
errx(1, "pipefs_nfsdir path name too long");
+ snprintf(pipefs_nfscbdir, sizeof(pipefs_nfscbdir), "%s/%s",
+ pipefs_dir, GSSD_NFSCB_NAME);
+ if (pipefs_nfscbdir[sizeof(pipefs_nfscbdir)-1] != '\0')
+ errx(1, "pipefs_nfscbdir path name too long");
if ((progname = strrchr(argv[0], '/')))
progname++;
diff --git a/utils/gssd/gssd.h b/utils/gssd/gssd.h
index 3c52f46..ef2ac54 100644
--- a/utils/gssd/gssd.h
+++ b/utils/gssd/gssd.h
@@ -49,6 +49,7 @@
#define GSSD_DEFAULT_MACHINE_CRED_SUFFIX "machine"
#define GSSD_DEFAULT_KEYTAB_FILE "/etc/krb5.keytab"
#define GSSD_SERVICE_NAME "nfs"
+#define GSSD_NFSCB_NAME "nfsd4_cb"
#define GSSD_SERVICE_NAME_LEN 3
#define GSSD_MAX_CCACHE_SEARCH 16
@@ -61,6 +62,7 @@ enum {AUTHTYPE_KRB5, AUTHTYPE_SPKM3, AUTHTYPE_LIPKEY};
extern char pipefs_dir[PATH_MAX];
extern char pipefs_nfsdir[PATH_MAX];
+extern char pipefs_nfscbdir[PATH_MAX];
extern char keytabfile[PATH_MAX];
extern char *ccachesearch[];
extern int use_memcache;
diff --git a/utils/gssd/gssd_main_loop.c b/utils/gssd/gssd_main_loop.c
index 397fd14..1c1ff4f 100644
--- a/utils/gssd/gssd_main_loop.c
+++ b/utils/gssd/gssd_main_loop.c
@@ -103,7 +103,7 @@ gssd_run()
{
int ret;
struct sigaction dn_act;
- int fd;
+ int fd, fd_cb;
sigset_t set;
/* Taken from linux/Documentation/dnotify.txt: */
@@ -125,6 +125,20 @@ gssd_run()
fcntl(fd, F_SETSIG, DNOTIFY_SIGNAL);
fcntl(fd, F_NOTIFY, DN_CREATE|DN_DELETE|DN_MODIFY|DN_MULTISHOT);
+ /* Attempt to open new callback pipe. If the open fails,
+ * don't try to process it. */
+ if ((fd_cb = open(pipefs_nfscbdir, O_RDONLY)) == -1) {
+ /* could be an older kernel or a newer one doing NFS 4.1 */
+ if (errno != ENOENT)
+ printerr(0, "WARNING: failed to open %s: %s\n",
+ pipefs_nfscbdir, strerror(errno));
+ memset(pipefs_nfscbdir, '\0', sizeof(pipefs_nfscbdir));
+ } else {
+ fcntl(fd_cb, F_SETSIG, DNOTIFY_SIGNAL);
+ fcntl(fd_cb, F_NOTIFY,
+ DN_CREATE|DN_DELETE|DN_MODIFY|DN_MULTISHOT);
+ }
+
init_client_list();
printerr(1, "beginning poll\n");
@@ -151,5 +165,7 @@ gssd_run()
}
}
close(fd);
+ if (fd_cb != -1)
+ close(fd_cb);
return;
}
diff --git a/utils/gssd/gssd_proc.c b/utils/gssd/gssd_proc.c
index 0fc0c42..969d113 100644
--- a/utils/gssd/gssd_proc.c
+++ b/utils/gssd/gssd_proc.c
@@ -232,11 +232,19 @@ read_service_info(char *info_file_name, char **servicename, char **servername,
sscanf(p, "port: %127s\n", cb_port);
/* check service, program, and version */
- if(memcmp(service, "nfs", 3)) return -1;
+ if (memcmp(service, "nfs", 3) != 0)
+ return -1;
*prog = atoi(program + 1); /* skip open paren */
*vers = atoi(version);
- if((*prog != 100003) || ((*vers != 2) && (*vers != 3) && (*vers != 4)))
- goto fail;
+
+ if (strlen(service) == 3 ) {
+ if ((*prog != 100003) || ((*vers != 2) && (*vers != 3) &&
+ (*vers != 4)))
+ goto fail;
+ } else if (memcmp(service, "nfs4_cb", 7) == 0) {
+ if (*vers != 1)
+ goto fail;
+ }
if (cb_port[0] != '\0') {
port = atoi(cb_port);
@@ -315,19 +323,18 @@ out:
static int
process_clnt_dir_files(struct clnt_info * clp)
{
- char kname[32];
- char sname[32];
- char info_file_name[32];
+ char name[PATH_MAX];
+ char info_file_name[PATH_MAX];
if (clp->krb5_fd == -1) {
- snprintf(kname, sizeof(kname), "%s/krb5", clp->dirname);
- clp->krb5_fd = open(kname, O_RDWR);
+ snprintf(name, sizeof(name), "%s/krb5", clp->dirname);
+ clp->krb5_fd = open(name, O_RDWR);
}
if (clp->spkm3_fd == -1) {
- snprintf(sname, sizeof(sname), "%s/spkm3", clp->dirname);
- clp->spkm3_fd = open(sname, O_RDWR);
+ snprintf(name, sizeof(name), "%s/spkm3", clp->dirname);
+ clp->spkm3_fd = open(name, O_RDWR);
}
- if((clp->krb5_fd == -1) && (clp->spkm3_fd == -1))
+ if ((clp->krb5_fd == -1) && (clp->spkm3_fd == -1))
return -1;
snprintf(info_file_name, sizeof(info_file_name), "%s/info",
clp->dirname);
@@ -384,17 +391,18 @@ insert_clnt_poll(struct clnt_info *clp)
}
static void
-process_clnt_dir(char *dir)
+process_clnt_dir(char *dir, char *pdir)
{
struct clnt_info * clp;
if (!(clp = insert_new_clnt()))
goto fail_destroy_client;
- if (!(clp->dirname = calloc(strlen(dir) + 1, 1))) {
+ /* An extra for the '/', and an extra for the null */
+ if (!(clp->dirname = calloc(strlen(dir) + strlen(pdir) + 2, 1))) {
goto fail_destroy_client;
}
- memcpy(clp->dirname, dir, strlen(dir));
+ sprintf(clp->dirname, "%s/%s", pdir, dir);
if ((clp->dir_fd = open(clp->dirname, O_RDONLY)) == -1) {
printerr(0, "ERROR: can't open %s: %s\n",
clp->dirname, strerror(errno));
@@ -438,16 +446,24 @@ init_client_list(void)
* directories, since the DNOTIFY could have been in there.
*/
static void
-update_old_clients(struct dirent **namelist, int size)
+update_old_clients(struct dirent **namelist, int size, char *pdir)
{
struct clnt_info *clp;
void *saveprev;
int i, stillhere;
+ char fname[PATH_MAX];
for (clp = clnt_list.tqh_first; clp != NULL; clp = clp->list.tqe_next) {
+ /* only compare entries in the global list that are from the
+ * same pipefs parent directory as "pdir"
+ */
+ if (strncmp(clp->dirname, pdir, strlen(pdir)) != 0) continue;
+
stillhere = 0;
for (i=0; i < size; i++) {
- if (!strcmp(clp->dirname, namelist[i]->d_name)) {
+ snprintf(fname, sizeof(fname), "%s/%s",
+ pdir, namelist[i]->d_name);
+ if (strcmp(clp->dirname, fname) == 0) {
stillhere = 1;
break;
}
@@ -468,13 +484,16 @@ update_old_clients(struct dirent **namelist, int size)
/* Search for a client by directory name, return 1 if found, 0 otherwise */
static int
-find_client(char *dirname)
+find_client(char *dirname, char *pdir)
{
struct clnt_info *clp;
+ char fname[PATH_MAX];
- for (clp = clnt_list.tqh_first; clp != NULL; clp = clp->list.tqe_next)
- if (!strcmp(clp->dirname, dirname))
+ for (clp = clnt_list.tqh_first; clp != NULL; clp = clp->list.tqe_next) {
+ snprintf(fname, sizeof(fname), "%s/%s", pdir, dirname);
+ if (strcmp(clp->dirname, fname) == 0)
return 1;
+ }
return 0;
}
@@ -497,12 +516,12 @@ process_pipedir(char *pipe_name)
return -1;
}
- update_old_clients(namelist, j);
+ update_old_clients(namelist, j, pipe_name);
for (i=0; i < j; i++) {
if (i < FD_ALLOC_BLOCK
&& !strncmp(namelist[i]->d_name, "clnt", 4)
- && !find_client(namelist[i]->d_name))
- process_clnt_dir(namelist[i]->d_name);
+ && !find_client(namelist[i]->d_name, pipe_name))
+ process_clnt_dir(namelist[i]->d_name, pipe_name);
free(namelist[i]);
}
@@ -510,7 +529,6 @@ process_pipedir(char *pipe_name)
return 0;
}
-
/* Used to read (and re-read) list of clients, set up poll array. */
int
update_client_list(void)
@@ -521,6 +539,15 @@ update_client_list(void)
if (retval)
printerr(0, "ERROR: processing %s\n", pipefs_nfsdir);
+ /* If we successfully processed nfsdir and callback directory exists
+ * process any events in the callback directory
+ */
+ if (retval == 0 && pipefs_nfscbdir[0] != '\0') {
+ retval = process_pipedir(pipefs_nfscbdir);
+ if (retval)
+ printerr(0, "ERROR: processing %s\n", pipefs_nfscbdir);
+ }
+
return retval;
}
Since I'm just getting around to this now... just
replace this patch please....
tia,
steved
Kevin Coffman wrote:
> On Thu, Apr 30, 2009 at 5:24 PM, J. Bruce Fields <[email protected]> wrote:
>> On Wed, Apr 29, 2009 at 05:56:26PM -0400, Kevin Coffman wrote:
>>> From: Olga Kornievskaia <[email protected]>
>>>
>>> Add support for handling upcalls on the new "nfsd4_cb" directory pipes.
>>> Only new kernels (2.6.29) have support for this new pipe directory.
>>> (The need for this new pipe directory will go away with NFSv4.1 where
>>> the callback can be done on the same connection as the fore-channel.)
>> My only complaint is that the code would be robust (and more
>> future-proof) if instead of specifically looking for "nfs" and
>> "nfsd4_cb", we just look at all top-level rpc_pipefs directories and
>> handed directories under any of them in the same way.
>>
>> --b.
>
> In our offline discussion, Bruce convinced me that we should just
> treat all the directories under the rpc_pipefs directory as equal, and
> process any clnt directories that show up within them. (This
> currently includes, "lockd mount nfs nfsd4_cb portmap statd".)
> Any new directories appearing in the future will automatically get the
> same treatment.
>
> Steve, I don't know what you might have already done with these
> patches. Would you prefer a replacement for this patch, or patch on
> top of this?
>
> K.C.
OK. I'll squash this in and re-submit the series on Wednesday.
On Mon, May 18, 2009 at 11:18 AM, Steve Dickson <[email protected]> wro=
te:
> Since I'm just getting around to this now... just
> replace this patch please....
>
> tia,
>
> steved
>
> Kevin Coffman wrote:
>> On Thu, Apr 30, 2009 at 5:24 PM, J. Bruce Fields <[email protected]=
rg> wrote:
>>> On Wed, Apr 29, 2009 at 05:56:26PM -0400, Kevin Coffman wrote:
>>>> From: Olga Kornievskaia <[email protected]>
>>>>
>>>> Add support for handling upcalls on the new "nfsd4_cb" directory p=
ipes.
>>>> Only new kernels (2.6.29) have support for this new pipe directory=
=2E
>>>> (The need for this new pipe directory will go away with NFSv4.1 wh=
ere
>>>> the callback can be done on the same connection as the fore-channe=
l.)
>>> My only complaint is that the code would be robust (and more
>>> future-proof) if instead of specifically looking for "nfs" and
>>> "nfsd4_cb", we just look at all top-level rpc_pipefs directories an=
d
>>> handed directories under any of them in the same way.
>>>
>>> --b.
>>
>> In our offline discussion, Bruce convinced me that we should just
>> treat all the directories under the rpc_pipefs directory as equal, a=
nd
>> process any clnt directories that show up within them. =A0(This
>> currently includes, "lockd =A0mount =A0nfs =A0nfsd4_cb =A0portmap =A0=
statd".)
>> Any new directories appearing in the future will automatically get t=
he
>> same treatment.
>>
>> Steve, I don't know what you might have already done with these
>> patches. =A0Would you prefer a replacement for this patch, or patch =
on
>> top of this?
>>
>> K.C.
>
>