staging/lustre: Replace jobid acquiring with per node setting
authorOleg Drokin <green@linuxhacker.ru>
Mon, 28 Apr 2014 02:21:52 +0000 (22:21 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 3 May 2014 23:33:15 +0000 (19:33 -0400)
Insted of meddling directly in process environment variables
(which is also not possible on certain platforms due to not exported
symbols), create jobid_name proc file to represent this info
(to be filled by job scheduler epilogue).

Signed-off-by: Oleg Drokin <oleg.drokin@intel.com>
CC: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/lustre/include/linux/libcfs/curproc.h
drivers/staging/lustre/lustre/include/lprocfs_status.h
drivers/staging/lustre/lustre/include/obd_class.h
drivers/staging/lustre/lustre/libcfs/linux/linux-curproc.c
drivers/staging/lustre/lustre/obdclass/class_obd.c
drivers/staging/lustre/lustre/obdclass/linux/linux-module.c

index 8fd47c98fd330eaffed0fe27b319a49fbe9759bc..b314f34d2e680def0de187f00eaa6c917b287be4 100644 (file)
@@ -56,7 +56,6 @@
 /* check if task is running in compat mode.*/
 #define current_pid()          (current->pid)
 #define current_comm()         (current->comm)
-int cfs_get_environ(const char *key, char *value, int *val_len);
 
 typedef __u32 cfs_cap_t;
 
index 9ce12c78d47a740e89fb405bf86c2c589c30e55d..1b7f6a9bf62cf13bae50354f986c91131c616e9a 100644 (file)
@@ -369,6 +369,7 @@ static inline void s2dhms(struct dhms *ts, time_t secs)
 #define JOBSTATS_JOBID_VAR_MAX_LEN     20
 #define JOBSTATS_DISABLE               "disable"
 #define JOBSTATS_PROCNAME_UID          "procname_uid"
+#define JOBSTATS_NODELOCAL             "nodelocal"
 
 extern int lprocfs_write_frac_helper(const char *buffer, unsigned long count,
                                     int *val, int mult);
index 61ba370794f8c8a1d92779f670397998ecb856d4..e265820c009fa3d8566383899716184755d20c0b 100644 (file)
@@ -2182,6 +2182,9 @@ void class_exit_uuidlist(void);
 int mea_name2idx(struct lmv_stripe_md *mea, const char *name, int namelen);
 int raw_name2idx(int hashtype, int count, const char *name, int namelen);
 
+/* class_obd.c */
+extern char obd_jobid_node[];
+
 /* prng.c */
 #define ll_generate_random_uuid(uuid_out) cfs_get_random_bytes(uuid_out, sizeof(class_uuid_t))
 
index e74c3e28a9727edc2e2a76f34a60e19fb008e744..bd301ce02255ee4ff273b0283ed13300133695fd 100644 (file)
@@ -100,158 +100,6 @@ cfs_cap_t cfs_curproc_cap_pack(void)
        return cap;
 }
 
-static int cfs_access_process_vm(struct task_struct *tsk, unsigned long addr,
-                                void *buf, int len, int write)
-{
-       /* Just copied from kernel for the kernels which doesn't
-        * have access_process_vm() exported */
-       struct mm_struct *mm;
-       struct vm_area_struct *vma;
-       struct page *page;
-       void *old_buf = buf;
-
-       mm = get_task_mm(tsk);
-       if (!mm)
-               return 0;
-
-       down_read(&mm->mmap_sem);
-       /* ignore errors, just check how much was successfully transferred */
-       while (len) {
-               int bytes, rc, offset;
-               void *maddr;
-
-               rc = get_user_pages(tsk, mm, addr, 1,
-                                    write, 1, &page, &vma);
-               if (rc <= 0)
-                       break;
-
-               bytes = len;
-               offset = addr & (PAGE_SIZE-1);
-               if (bytes > PAGE_SIZE-offset)
-                       bytes = PAGE_SIZE-offset;
-
-               maddr = kmap(page);
-               if (write) {
-                       copy_to_user_page(vma, page, addr,
-                                         maddr + offset, buf, bytes);
-                       set_page_dirty_lock(page);
-               } else {
-                       copy_from_user_page(vma, page, addr,
-                                           buf, maddr + offset, bytes);
-               }
-               kunmap(page);
-               page_cache_release(page);
-               len -= bytes;
-               buf += bytes;
-               addr += bytes;
-       }
-       up_read(&mm->mmap_sem);
-       mmput(mm);
-
-       return buf - old_buf;
-}
-
-/* Read the environment variable of current process specified by @key. */
-int cfs_get_environ(const char *key, char *value, int *val_len)
-{
-       struct mm_struct *mm;
-       char *buffer, *tmp_buf = NULL;
-       int buf_len = PAGE_CACHE_SIZE;
-       int key_len = strlen(key);
-       unsigned long addr;
-       int rc;
-
-       buffer = kmalloc(buf_len, GFP_USER);
-       if (!buffer)
-               return -ENOMEM;
-
-       mm = get_task_mm(current);
-       if (!mm) {
-               kfree(buffer);
-               return -EINVAL;
-       }
-
-       /* Avoid deadlocks on mmap_sem if called from sys_mmap_pgoff(),
-        * which is already holding mmap_sem for writes.  If some other
-        * thread gets the write lock in the meantime, this thread will
-        * block, but at least it won't deadlock on itself.  LU-1735 */
-       if (down_read_trylock(&mm->mmap_sem) == 0) {
-               kfree(buffer);
-               return -EDEADLK;
-       }
-       up_read(&mm->mmap_sem);
-
-       addr = mm->env_start;
-       while (addr < mm->env_end) {
-               int this_len, retval, scan_len;
-               char *env_start, *env_end;
-
-               memset(buffer, 0, buf_len);
-
-               this_len = min_t(int, mm->env_end - addr, buf_len);
-               retval = cfs_access_process_vm(current, addr, buffer,
-                                              this_len, 0);
-               if (retval != this_len)
-                       break;
-
-               addr += retval;
-
-               /* Parse the buffer to find out the specified key/value pair.
-                * The "key=value" entries are separated by '\0'. */
-               env_start = buffer;
-               scan_len = this_len;
-               while (scan_len) {
-                       char *entry;
-                       int entry_len;
-
-                       env_end = memscan(env_start, '\0', scan_len);
-                       LASSERT(env_end >= env_start &&
-                               env_end <= env_start + scan_len);
-
-                       /* The last entry of this buffer cross the buffer
-                        * boundary, reread it in next cycle. */
-                       if (unlikely(env_end - env_start == scan_len)) {
-                               /* This entry is too large to fit in buffer */
-                               if (unlikely(scan_len == this_len)) {
-                                       CERROR("Too long env variable.\n");
-                                       GOTO(out, rc = -EINVAL);
-                               }
-                               addr -= scan_len;
-                               break;
-                       }
-
-                       entry = env_start;
-                       entry_len = env_end - env_start;
-
-                       /* Key length + length of '=' */
-                       if (entry_len > key_len + 1 &&
-                           !memcmp(entry, key, key_len)) {
-                               entry += key_len + 1;
-                               entry_len -= key_len + 1;
-                               /* The 'value' buffer passed in is too small.*/
-                               if (entry_len >= *val_len)
-                                       GOTO(out, rc = -EOVERFLOW);
-
-                               memcpy(value, entry, entry_len);
-                               *val_len = entry_len;
-                               GOTO(out, rc = 0);
-                       }
-
-                       scan_len -= (env_end - env_start + 1);
-                       env_start = env_end + 1;
-               }
-       }
-       GOTO(out, rc = -ENOENT);
-
-out:
-       mmput(mm);
-       kfree((void *)buffer);
-       if (tmp_buf)
-               kfree((void *)tmp_buf);
-       return rc;
-}
-EXPORT_SYMBOL(cfs_get_environ);
-
 EXPORT_SYMBOL(cfs_cap_raise);
 EXPORT_SYMBOL(cfs_cap_lower);
 EXPORT_SYMBOL(cfs_cap_raised);
index c93131e0d2dada3c9c2278c043d4aec8dff9fc6e..dde04b767a6d5a2ec331574ff66c153e2c4f8657 100644 (file)
@@ -102,23 +102,17 @@ EXPORT_SYMBOL(obd_dirty_transit_pages);
 char obd_jobid_var[JOBSTATS_JOBID_VAR_MAX_LEN + 1] = JOBSTATS_DISABLE;
 EXPORT_SYMBOL(obd_jobid_var);
 
-/* Get jobid of current process by reading the environment variable
- * stored in between the "env_start" & "env_end" of task struct.
- *
- * TODO:
- * It's better to cache the jobid for later use if there is any
- * efficient way, the cl_env code probably could be reused for this
- * purpose.
+char obd_jobid_node[JOBSTATS_JOBID_SIZE + 1];
+
+/* Get jobid of current process from stored variable or calculate
+ * it from pid and user_id.
  *
- * If some job scheduler doesn't store jobid in the "env_start/end",
- * then an upcall could be issued here to get the jobid by utilizing
- * the userspace tools/api. Then, the jobid must be cached.
+ * Historically this was also done by reading the environment variable
+ * stored in between the "env_start" & "env_end" of task struct.
+ * This is now deprecated.
  */
 int lustre_get_jobid(char *jobid)
 {
-       int jobid_len = JOBSTATS_JOBID_SIZE;
-       int rc = 0;
-
        memset(jobid, 0, JOBSTATS_JOBID_SIZE);
        /* Jobstats isn't enabled */
        if (strcmp(obd_jobid_var, JOBSTATS_DISABLE) == 0)
@@ -132,31 +126,13 @@ int lustre_get_jobid(char *jobid)
                return 0;
        }
 
-       rc = cfs_get_environ(obd_jobid_var, jobid, &jobid_len);
-       if (rc) {
-               if (rc == -EOVERFLOW) {
-                       /* For the PBS_JOBID and LOADL_STEP_ID keys (which are
-                        * variable length strings instead of just numbers), it
-                        * might make sense to keep the unique parts for JobID,
-                        * instead of just returning an error.  That means a
-                        * larger temp buffer for cfs_get_environ(), then
-                        * truncating the string at some separator to fit into
-                        * the specified jobid_len.  Fix later if needed. */
-                       static bool printed;
-                       if (unlikely(!printed)) {
-                               LCONSOLE_ERROR_MSG(0x16b, "%s value too large "
-                                                  "for JobID buffer (%d)\n",
-                                                  obd_jobid_var, jobid_len);
-                               printed = true;
-                       }
-               } else {
-                       CDEBUG((rc == -ENOENT || rc == -EINVAL ||
-                               rc == -EDEADLK) ? D_INFO : D_ERROR,
-                              "Get jobid for (%s) failed: rc = %d\n",
-                              obd_jobid_var, rc);
-               }
+       /* Whole node dedicated to single job */
+       if (strcmp(obd_jobid_var, JOBSTATS_NODELOCAL) == 0) {
+               strcpy(jobid, obd_jobid_node);
+               return 0;
        }
-       return rc;
+
+       return -ENOENT;
 }
 EXPORT_SYMBOL(lustre_get_jobid);
 
index 0334882070ef33f558ef0f88c15b5de22eeca13f..bdf2eed2952ab01fbe81809657ee26bf8de5cbfc 100644 (file)
@@ -292,6 +292,31 @@ static ssize_t obd_proc_jobid_var_seq_write(struct file *file, const char *buffe
 }
 LPROC_SEQ_FOPS(obd_proc_jobid_var);
 
+static int obd_proc_jobid_name_seq_show(struct seq_file *m, void *v)
+{
+       return seq_printf(m, "%s\n", obd_jobid_var);
+}
+
+static ssize_t obd_proc_jobid_name_seq_write(struct file *file,
+                                            const char __user *buffer,
+                                            size_t count, loff_t *off)
+{
+       if (!count || count > JOBSTATS_JOBID_SIZE)
+               return -EINVAL;
+
+       if (copy_from_user(obd_jobid_node, buffer, count))
+               return -EFAULT;
+
+       obd_jobid_node[count] = 0;
+
+       /* Trim the trailing '\n' if any */
+       if (obd_jobid_node[count - 1] == '\n')
+               obd_jobid_node[count - 1] = 0;
+
+       return count;
+}
+LPROC_SEQ_FOPS(obd_proc_jobid_name);
+
 /* Root for /proc/fs/lustre */
 struct proc_dir_entry *proc_lustre_root = NULL;
 EXPORT_SYMBOL(proc_lustre_root);
@@ -301,6 +326,8 @@ struct lprocfs_vars lprocfs_base[] = {
        { "pinger", &obd_proc_pinger_fops },
        { "health_check", &obd_proc_health_fops },
        { "jobid_var", &obd_proc_jobid_var_fops },
+       { .name =       "jobid_name",
+         .fops =       &obd_proc_jobid_name_fops},
        { 0 }
 };