In-tree documentation improvements (LU-1892)

[LU-1914] Doxygen comments - ldlm module Created: 12/Sep/12  Updated: 29/Apr/14  Resolved: 29/Apr/14

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: Lustre 2.4.0

Type: Technical task Priority: Minor
Reporter: Ned Bass Assignee: Ned Bass
Resolution: Fixed Votes: 0
Labels: documentation

Attachments: File ldlm-doxygen-3.diff    
Bugzilla ID: 20,943
Rank (Obsolete): 8140

 Description   

Update the attached patch from Bug 20943, which adds doxygen comments to the ldlm module, for master.



 Comments   
Comment by Ned Bass [ 12/Sep/12 ]

I have a few questions about this patch.

         /**
-         * Hash table for namespace.
+         * Resource hash table for namespace.
+         * Hash table is organized as an array of \see RES_HASH_SIZE elements.
+         * Each element is a list_head linking all resources whose name hash is
+         * equal to the array position.
+         * \see ldlm_hash_fn for the actual hash function.
          */
         cfs_list_t            *ns_hash;

The ldlm namespace resource hashtable now appears to use the generic cfs hash implementation, so I think we can omit any hash table implementation details here.

-         /**
-          * Seconds.
-          */
+        /**
+         * Seconds.
+         * Appears to be currently unused. Should be removed.
+         */
         unsigned int           ns_ctime_age_limit;

Contrary to this comment, ns_ctime_age_limit is still used in mdt_handler.c:mdt_getattr_name_lock() (bugzilla 14910). I'm thinking of of a comment along these lines.

/** 
 * Number of seconds since the file change time after which the 
 * MDT will return an UPDATE lock along with a LOOKUP lock. 
 * This allows the client to start caching negative dentries 
 * for the directory and may save an RPC for a later stat. 
 */ 
unsigned int           ns_ctime_age_limit;

But it's still a bit muddled to me. What is the connection between the change time, the update lock, negative dentries, and why does returning an update lock save an RPC for a stat? How is the value chosen? This may not be the right place to explain all these things; perhaps the LDLM_CTIME_AGE_LIMIT definition would be appropriate, and we could provide a reference here.

Comment by Ned Bass [ 13/Sep/12 ]

http://review.whamcloud.com/#change,3985

Comment by Ned Bass [ 18/Dec/12 ]

patch landed to master

Generated at Sat Feb 10 01:20:47 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.