Re: [PATCH] 2.6: improved fdmap

[Hans Reiser <reiser@namesys.com>]

There is some commonality of concepts between this and reiser4 flows.  I 
thought you might find that of interest.

Hans

Matt Miller wrote:

>+/*
>+ * purpose
>+ *
>+ *    Map process memory ranges to virtual file descriptors.  This can be
>+ *    thought of as being the opposite of mmap, simply, instead of mapping
>+ *    a file's contents to a memory range, fdmap maps a memory range to
>+ *    a virtual file.  This allows one to read, write, seek, and even
>+ *    mmap the virtual file's contents as if it were a real file on disk.
>+ *
>+ * interface
>+ *
>+ *    fdmap exposes a new system call identifier.  The system call takes
>+ *    arguments as the following prototype conveys:
>+ *
>+ *       int fdmap(void *addr, size_t len, int flags);
>+ *
>+ *    ``flags'' can be one of O_RDONLY, O_WRONLY, or O_RDWR.
>+ *
>+ *    syscall number
>+ *
>+ *       alpha, arm, ia32, sh: 274
>+ *       sparc, sparc64: 218
>+ *       ia64: 1259
>+ *       m68k: 236
>+ *       mips: 32b=4268 64b=5227 64be32=6231
>+ *       parisc: 229
>+ *       ppc, ppc64: 256
>+ *       s390: 265
>+ *       v850: 203
>+ *
>+ * based on
>+ *
>+ *    The underlying virtual filesystem code was adapted from sockfs.
>+ *
>+ * Matt Miller
>+ * mmiller@hick.org
>+ */
>  
>


-- 
Hans