Changeset 626 for trunk/kernel/fs/vfs.h

Timestamp:

Apr 29, 2019, 7:25:09 PM (5 years ago)

Author:

alain

Message:

This version has been tested on the sort multithreaded application
for TSAR_IOB architectures ranging from 1 to 8 clusters.
It fixes three bigs bugs:
1) the dev_ioc device API has been modified: the dev_ioc_sync_read()
and dev_ioc_sync_write() function use now extended pointers on the
kernel buffer to access a mapper stored in any cluster.
2) the hal_uspace API has been modified: the hal_copy_to_uspace()
and hal_copy_from_uspace() functions use now a (cxy,ptr) couple
to identify the target buffer (equivalent to an extended pointer.
3) an implementation bug has been fixed in the assembly code contained
in the hal_copy_to_uspace() and hal_copy_from_uspace() functions.

File:

• trunk/kernel/fs/vfs.h (modified) (4 diffs)

trunk/kernel/fs/vfs.h

@@ -376 +376 @@
  * argument to a local buffer identified by the <name> argument.
  * The local buffer size must be at least CONFIG_VFS_MAX_NAME_LENGTH. 
- *****************************************************************************************
+ ******************************************************************************************
  * @ inode_xp  : extended pointer on the remote inode.
  * @ name      : local buffer pointer.
@@ -396 +396 @@
 error_t vfs_inode_load_all_pages( vfs_inode_t * inode );
 
-
+/****************************************************************************************** 
+ * This debug function display the curren state of an inode descriptor identified by
+ * the <inode_xp> argument.
+ *****************************************************************************************/
+void vfs_inode_display( xptr_t inode_xp );
 
 /******************************************************************************************
@@ -547 +551 @@
  * It can be executed by any thread running in any cluster (can be different from both
  * the child cluster and the parent cluster).
- * 
- * [Implementation]
+ * The new child inode and the parent inode can have different FS types.
+ * [Implementation note]
  * As there are cross-references between inode and dentry, this function implements
- * a three steps scenario :
+ * a five steps scenario :
  * 1) The dentry descriptor is created in the cluster containing the existing <parent_xp>
  *    inode, and partially initialized, using the RPC_VFS_CREATE DENTRY if required.
  * 2) The inode and its associated mapper are created in cluster identified by <child_cxy>,
  *    and partially initialised, using the RPC_VFS_CREATE_INODE if required.
- *    The new inode and the parent inode can have different FS types.
- * 3) The pointers between the parent inode, the new dentry, and the child inode 
- *    are updated, using remote accesses.
- ******************************************************************************************
+ * 3) The pointers on dentry in parent inode are updated, using remote access.
+ * 4) The pointers on dentry in child inode are updated, using remotes access.
+ * 5) The pointers on parent and child inode in dentry are updated, using remotes access.
+ *****************************************************************************************
  * @ child_inode_cxy  : [in]  target cluster for child inode.
  * @ fs_type          : [in]  child inode FS type.
@@ -591 +595 @@
 
 /******************************************************************************************
- * This function is called by the vfs_lookup() function when a new dentry/inode must
- * be created from scratch and introduced in both the Inode Tree and the IOC device.
- * The dentry and inode descriptors have been created by the caller.
- * - It allocates one cluster from the relevant FS, and updates the File Allocation
- *   Table (both the FAT mapper, and the IOC device).
- * - It set the "size", and "extend" fields in child inode descriptor.
- * - It updates the parent directory to introduce the new child in the parent directory
- *   inode descriptor (radix tree), in theparent inode mapper, and on IOC device.
- * - It set the "extend" field in dentry descriptor.
+ * This function is called by the vfs_lookup() function when a new (dentry/inode) must
+ * be created from scratch and introduced in both the parent mapper and the IOC device.
+ * The dentry and inode descriptors must have been previously created by the caller.
+ * 1. It allocates one cluster from the relevant FS, updates the FAT mapper,
+ *    and synchronously update the IOC device).
+ * 2. It set the "size", and "extend" fields in child inode descriptor.
+ * 3. It updates the parent directory mapper to introduce the new child,
+ *    and synchronously update the IOC device.
+ * 4. It set the "extend" field in dentry descriptor.
  * It can be called by a thread running in any cluster.
  ******************************************************************************************