Dynamically configurable reverse DNLC lookup

• US 7,822,927 B1
• Filed: 05/14/2007
• Issued: 10/26/2010
• Est. Priority Date: 05/14/2007
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Scheer, Michael D.
• Primary Examiner(s): Peugh; Brian R
• Assistant Examiner(s): SIMONETTI, NICHOLAS J
• Application Number:
  US11/748,457
• Time in Patent Office: 1,261 Days
• Field of Search: 711/205, 711/216
• US Class Current: 711/133
• CPC Class Codes: G06F 12/0866 for peripheral storage systems, e.g. disk cache
• CPC Class Codes: G06F 2212/466 Metadata, control data

Abstract

A directory name lookup cache (DNLC) provides a hashed forward mapping for finding the “child handle” associated with a “parent handle” and a “child name.” To provide an efficient reverse lookup capability, a second set of links is added to each cache entry for a “child hash list” indexed by a hashing of the child handle. For dynamically enabling and disabling the reverse mapping, when a new cache entry is added to its parent hash list, if the reverse mapping is enabled, then the new cache entry is also added to its child hash list; otherwise, the new cache entry is marked to indicate that it is not in any child hash list. To save memory, the parent hash lists and the child hash lists may share hash buckets.

Claims

1. A method of operation of a data processing system including an electronic random access memory and at least one data processor coupled to the random access memory for writing data to and reading data from the random access memory, said method comprising:

storing cache entries in the random access memory, each of the cache entries including a parent handle, a child handle, and an alphanumeric child name, the parent handle being a handle of a directory in a file system, the child handle being a handle of a subdirectory or file in the directory, and the child name being a name of the subdirectory or file;

preempting cache entries that are not frequently accessed for storing new cache entries in the random access memory;

maintaining parent hash lists of the cache entries and searching the random access memory for a child handle associated with a specified parent handle and a specified child name by searching a parent hash list indexed by a hashing of the specified parent handle and the specified child name;

maintaining child hash lists of the cache entries and searching the random access memory for a parent handle and a child name associated with a specified child handle by searching a child hash list indexed by a hashing of the specified child handle; and

inserting the new cache entries into the parent hash lists, and dynamically enabling and disabling insertion of the new cache entries into the child hash lists, so that when the insertion of the new cache entries into the child hash lists is enabled, new cache entries are inserted into the parent hash lists and are also inserted into the child hash lists, and when the insertion of the new cache entries into the child hash lists is disabled, new cache entries having child handles are inserted into the parent hash lists and are not inserted into any of the child hash lists.

2. The method as claimed in claim 1, which further includes marking as not being in any child hash list at least one new cache entry that has a child handle and is inserted into a parent hash list when insertion of the new cache entries into the child hash lists is disabled.

3. The method as claimed in claim 2, which includes marking said at least one new cache entry as not being in any child hash list by setting a child hash list pointer field in said at least one new cache entry to a value that is invalid for a child hash list pointer.

4. The method as claimed in claim 2, wherein the preempting of a not frequently accessed cache entry includes obtaining a lock on a parent hash list of the not frequently accessed cache entry, inspecting the not frequently accessed cache entry to determine that the not frequently accessed cache entry is not marked as not being in any child hash list, and upon determining that the not frequently accessed cache entry is not marked as not being in any child hash list, obtaining a lock on a child hash list of the not frequently accessed cache entry and removing the not frequently accessed cache entry from the child hash list of the not frequently accessed cache entry.

5. The method as claimed in claim 1, wherein the parent hash lists and the child hash lists share hash buckets and locks upon the hash buckets, and wherein the method further includes avoiding a deadlock when execution of a first code thread has obtained a lock upon one hash bucket for accessing a parent hash list anchored to said one hash bucket and the first code thread is seeking a lock upon another hash bucket for accessing a child hash list anchored to said another hash bucket, and execution of a second code thread has obtained the lock upon said another hash bucket for accessing a parent hash list anchored to said another hash bucket and the second code thread is seeking the lock upon said one hash bucket for accessing a child hash list anchored to said one hash bucket.

6. The method as claimed in claim 5, which includes avoiding the deadlock by giving priority to one of the code threads based on a comparison of a hash index value of said one hash bucket to a hash index value of said another hash bucket, and by relinquishing the lock obtained by the other one of the code threads.

7. A data processing system comprising:

an electronic random access memory; and

at least one data processor coupled to the random access memory for writing data to and reading data from the random access memory;

wherein said at least one data processor is programmed for storing cache entries in the random access memory, each of the cache entries including a parent handle, a child handle, and an alphanumeric child name, the parent handle being a handle of a directory in a file system, the child handle being a handle of a subdirectory or file in the directory, and the child name being a name of the subdirectory or file;

wherein said at least one data processor is programmed for preempting cache entries that are not frequently accessed for storing new cache entries in the random access memory;

wherein said at least one data processor is programmed for maintaining parent hash lists of the cache entries and searching the random access memory for a child handle associated with a specified parent handle and a specified child name by searching a parent hash list indexed by a hashing of the specified parent handle and the specified child name;

wherein said at least one data processor is programmed for maintaining child hash lists of the cache entries and searching the random access memory for a parent handle and a child name associated with a specified child handle by searching a child hash list indexed by a hashing of the specified child handle; and

wherein said at least one data processor is programmed for inserting the new cache entries into the parent hash lists, and for dynamically enabling and disabling insertion of the new cache entries into the child hash lists so that when the insertion of the new cache entries into the child hash lists is enabled, new cache entries are inserted into the parent hash lists and are also inserted into the child hash lists, and when the insertion of the new cache entries into the child hash lists is disabled, new cache entries having child handles are inserted into the parent hash lists and are not inserted into any of the child hash lists.

8. The data processing system as claimed in claim 7, wherein said at least one data processor is further programmed for marking a new cache entry having a child handle as not being in any child hash list when the new cache entry is inserted into the parent hash list and insertion of the new cache entries into the child hash lists is disabled.

9. The data processing system as claimed in claim 8, wherein said at least one data processor is further programmed for marking the new cache entry as not being in any child hash list by setting a child hash list pointer field in the new cache entry to a value that is invalid for a child hash list pointer.

10. The data processing system as claimed in claim 8, wherein said at least one data processor is programmed for preempting a not frequently accessed cache entry by obtaining a lock on a parent hash list of the not frequently accessed cache entry, inspecting the not frequently accessed cache entry to determine whether the not frequently accessed cache entry is marked as not being in any child hash list, and if the not frequently accessed cache entry is marked as not being in any child hash list, then removing the not frequently accessed cache entry from the parent hash list of the not frequently accessed cache entry, and otherwise if the not frequently accessed cache entry is not marked as not being in any child hash list, then obtaining a lock on a child hash list of the not frequently accessed cache entry and removing the not frequently accessed cache entry from the child hash list of the not frequently accessed cache entry and removing the not frequently accessed cache entry from the parent hash list of the not frequently accessed cache entry.

11. The data processing system as claimed in claim 7, wherein said at least one data processor is programmed for maintaining a parent hash table of pointers to the parent hash lists and locks upon the parent hash lists and for indexing the parent hash table with the hashing of the specified parent handle and the specified child name when searching the cache for the child handle associated with the specified parent handle and the specified child name;

and

wherein said at least one data processor is programmed for maintaining a child hash table of pointers to the child hash lists and locks upon the child hash lists and for indexing the child hash table with the hashing of the specified child handle when searching the cache for the parent handle and child name associated with the specified child handle.

12. The data processing system as claimed in claim 7, wherein said at least one data processor is programmed for maintaining a hash table of pointers to the parent hash lists and the child hash lists and locks shared upon the parent hash lists the child hash lists, and for indexing the hash table with the hashing of the specified parent handle and the specified child name when searching the cache for the child handle associated with the specified parent handle and the specified child name, and for indexing the hash table with the hashing of the specified child handle when searching the cache for the parent handle and child name associated with the specified child handle.

13. The data processing system as claimed in claim 12, wherein said at least one data processor is programmed for avoiding a deadlock when execution of a first code thread has obtained a lock of one row of the hash table for accessing a parent hash list anchored in said one row of the hash table and the first code thread is seeking a lock of another row of the hash table for accessing a child hash list anchored in said another row of the hash table, and execution of a second code thread has obtained the lock of said another row of the hash table for accessing a parent hash list anchored in said another row of the hash table and the second code thread is seeking the lock of said one row of the hash table for accessing a child hash list anchored in said one row of the hash table.

14. The data processing system as claimed in claim 13, wherein said at least one data processor is further programmed for avoiding the deadlock by giving priority to one of the code threads based on a comparison of a hash table index of said one row of the hash table to a hash table index of said another row of the hash table, and relinquishing the lock obtained by the other one of the code threads.

15. A file server comprising, in combination:

data storage for storing a file system;

electronic random access memory; and

at least one data processor coupled to the data storage for providing a client with access to the file system, and coupled to the random access memory for writing data to and reading data from the random access memory;

wherein said at least one data processor is programmed for storing cache entries in the random access memory, each of the cache entries including a parent handle, a child handle, and an alphanumeric child name, the parent handle being a handle of a directory in the file system, the child handle being a handle of a subdirectory or file in the directory, and the child name being a name of the subdirectory or file;

wherein said at least one data processor is programmed for preempting cache entries that are not frequently accessed for storing new cache entries in the random access memory;

wherein said at least one data processor is programmed for maintaining parent hash lists of the cache entries and searching the random access memory for a child handle associated with a specified parent handle and a specified child name by searching a parent hash list indexed by a hashing of the specified parent handle and the specified child name;

wherein said at least one data processor is programmed for maintaining child hash lists of the cache entries and searching the random access memory for a parent handle and a child name associated with a specified child handle by searching a child hash list indexed by a hashing of the specified child handle;

wherein said at least one data processor is programmed for responding to a request from the client for finding a file having a specified name in the file system by performing a forward lookup in the random access memory for a file handle of the file having the specified name and returning the file handle to the client;

wherein said at least one data processor is programmed for responding to a request from the client for reading data from a file having a specified file handle by detecting an error upon finding that the file having the specified file handle is inaccessible from the data storage;

wherein said at least one data processor is programmed for performing a reverse lookup in the random access memory to determine a pathname in the file system for the file having the specified file handle for recovery from the error; and

wherein said at least one data processor is programmed for inserting the new cache entries into the parent hash lists, and for dynamically enabling and disabling insertion of the new cache entries into the child hash lists, so that when the insertion of the new cache entries into the child hash lists is enabled, new cache entries are inserted into the parent hash lists and are also inserted into the child hash lists, and when the insertion of the new cache entries into the child hash lists is disabled, new cache entries having child handles are inserted into the parent hash lists and are not inserted into any of the child hash lists.

16. The file server as claimed in claim 15, wherein said at least one data processor is further programmed for marking a new cache entry having a child handle as not being in any child hash list when the new cache entry is inserted into a parent hash list and insertion of the new cache entries into the child hash lists is disabled.

17. The file server as claimed in claim 15, wherein the parent hash lists and the child hash lists share hash buckets and locks upon the hash buckets, and wherein said at least one data processor is programmed for avoiding a deadlock when execution of a first code thread has obtained a lock upon one hash bucket for accessing a parent hash list anchored to said one hash bucket and the first code thread is seeking a lock upon another hash bucket for accessing a child hash list anchored to said another hash bucket, and execution of a second code thread has obtained the lock upon said another hash bucket for accessing a parent hash list anchored to said another hash bucket and the second code thread is seeking the lock upon said one hash bucket for accessing a child hash list anchored to said one hash bucket.