smbfs / loop: problematic or not unuseable?
Hello!
While still suffering from my more or less dead harddisk I am seeing other nice problems. I copied my blowfish encrypted partition to a samba server. Now I want to mount it and use it:
pc-it-nico# mount
[...]
//fs2/home on /home/nico/fs2 type smbfs (rw)
pc-it-nico# mount /home/nico/fs2/home-13-Feb-2004.tar.crypt /tmp/ -o loop,encryption=blowfish Password:
ioctl: LOOP_SET_FD: Invalid argument
Well, let's try to use a local copy:
pc-it-nico# cd ~nico/fs2
pc-it-nico# cp home-13-Feb-2004.tar.crypt ..
pc-it-nico# cd ..
pc-it-nico# mount -t xfs home-13-Feb-2004.tar.crypt scice -o loop,encryption=blowfish pc-it-nico#
This works fine.
Just wanted to report this and ask whether this is what it should be.
Greetings,
Nico
ps: If somebody can recommened me a good manufacturer for 2.5" notebook
disks, please send me a private mail. I am looking for something running longer than 2 months, having some performance, still beiing not too loud :)
Quoted 6 lines · Nico Schottelius
Hello!
While still suffering from my more or less dead harddisk I am seeing other nice problems. I copied my blowfish encrypted partition to a samba server. Now I want to mount it and use it:
Quoted 7 lines
pc-it-nico# mount
[...]
//fs2/home on/home/nico/fs2type smbfs (rw)pc-it-nico# mount
/home/nico/fs2/home-13-Feb-2004.tar.crypt/tmp/-o loop,encryption=blowfish Password:
ioctl:LOOP_SET_FD: Invalid argument
Yes, it is known. You can try the patch below.
I don't know if there is anything more to supporting sendfile than this. smb_file_sendfile follows the same pattern as the other smb_file_* operations.
/Urban
fs/smbfs/file.c +22 -0
@@ -257,6 +257,27 @@return status; } +static ssize_t +smb_file_sendfile (struct file *file,loff_t *ppos, +size_t count,read_actor_t actor,void __user *target) +{ +struct dentry *dentry = file->f_dentry; +ssize_t status; + +VERBOSE ("file %s/%s, pos=%Ld, count=%d\n" , +DENTRY_PATH (dentry), *ppos, count); + + status =smb_revalidate_inode (dentry); +if (status) { +PARANOIA ("%s/%s validation failed, error=%d\n" , +DENTRY_PATH (dentry), status); + goto out; + } + status =generic_file_sendfile (file, ppos, count, actor, target); +out: +return status; +} +/* * This does the "real" work of the write. The generic routine has * allocated the page, locked it, done all the page alignment stuff
@@ -388,6 +409,7 @@ .open = smb_file_open, .release = smb_file_release, .fsync = smb_fsync, + .sendfile = smb_file_sendfile, }; struct inode_operations smb_file_inode_operations =
fs/smbfs/proc.c +0 -6
@@ -1015,12 +1015,6 @@ p +=19 ; p +=8 ; -/* FIXME: the request will fail if the 'tid' is changed. This -should perhaps be set just before transmitting ... */ -WSET (req->rq_header, smb_tid, server->opt.tid); -WSET (req->rq_header, smb_pid,1 ); -WSET (req->rq_header, smb_uid, server->opt.server_uid); -if (server->opt.protocol >SMB_PROTOCOL_CORE ) {int flags =SMB_FLAGS_CASELESS_PATHNAMES ; int flags2 = SMB_FLAGS2_LONG_PATH_COMPONENTS |
fs/smbfs/request.c +6 -0
@@ -384,6 +384,12 @@struct smb_sb_info *server = req->rq_server;int result; +if (req->rq_bytes_sent ==0 ) { +WSET (req->rq_header, smb_tid, server->opt.tid); +WSET (req->rq_header, smb_pid,1 ); +WSET (req->rq_header, smb_uid, server->opt.server_uid); + } + result =smb_send_request (req);if (result <0 && result != -EAGAIN ) goto out;
fs/smbfs/smbiod.c +3 -0
@@ -161,6 +161,9 @@while (head != &server->xmitq) { req = list_entry(head, struct smb_request, rq_queue); head = head->next; + +WSET (req->rq_header, smb_uid,-1 ); + req->rq_bytes_sent =0 ;if (req->rq_flags &SMB_REQ_NORETRY ) {VERBOSE ("aborting request %p on xmitq\n" , req); req->rq_errno = -EIO ;
Works!
Will it be included into the main tree?
Nico
Urban Widmark [Mon, Feb 16, 2004 at 03:53:28PM +0100]:
Quoted 110 lines · Nico Schottelius
Quoted 7 lines
Hello!
While still suffering from my more or less dead harddisk I am seeing other nice problems. I copied my blowfish encrypted partition to a samba server. Now I want to mount it and use it:
Quoted 7 lines
pc-it-nico# mount
[...]
//fs2/home on/home/nico/fs2type smbfs (rw)pc-it-nico# mount
/home/nico/fs2/home-13-Feb-2004.tar.crypt/tmp/-o loop,encryption=blowfish Password:
ioctl:LOOP_SET_FD: Invalid argumentYes, it is known. You can try the patch below.
I don't know if there is anything more to supporting sendfile than this.
smb_file_sendfilefollows the same pattern as the other smb_file_* operations./Urban
diff -urN -X exclude
linux-2.6.3-rc1-orig/fs/smbfs/file.clinux-2.6.3-rc1-smbfs/fs/smbfs/file.cfs/smbfs/file.c +22
-0@@ -257,6 +257,27 @@return status; } +static ssize_t +smb_file_sendfile (struct file *file,loff_t *ppos, +size_t count,read_actor_t actor,void __user *target) +{ +struct dentry *dentry = file->f_dentry; +ssize_t status; + +VERBOSE ("file %s/%s, pos=%Ld, count=%d\n" , +DENTRY_PATH (dentry), *ppos, count); + + status =smb_revalidate_inode (dentry); +if (status) { +PARANOIA ("%s/%s validation failed, error=%d\n" , +DENTRY_PATH (dentry), status); + goto out; + } + status =generic_file_sendfile (file, ppos, count, actor, target); +out: +return status; +} +/* * This does the "real" work of the write. The generic routine has * allocated the page, locked it, done all the page alignment stuff @@ -388,6 +409,7 @@ .open = smb_file_open, .release = smb_file_release, .fsync = smb_fsync, + .sendfile = smb_file_sendfile, }; struct inode_operations smb_file_inode_operations =diff -urN -X exclude
linux-2.6.3-rc1-orig/fs/smbfs/proc.clinux-2.6.3-rc1-smbfs/fs/smbfs/proc.cfs/smbfs/proc.c +0
-6@@ -1015,12 +1015,6 @@ p +=19 ; p +=8 ; -/* FIXME: the request will fail if the 'tid' is changed. This -should perhaps be set just before transmitting ... */ -WSET (req->rq_header, smb_tid, server->opt.tid); -WSET (req->rq_header, smb_pid,1 ); -WSET (req->rq_header, smb_uid, server->opt.server_uid); -if (server->opt.protocol >SMB_PROTOCOL_CORE ) {int flags =SMB_FLAGS_CASELESS_PATHNAMES ; int flags2 = SMB_FLAGS2_LONG_PATH_COMPONENTS |diff -urN -X exclude
linux-2.6.3-rc1-orig/fs/smbfs/request.clinux-2.6.3-rc1-smbfs/fs/smbfs/request.cfs/smbfs/request.c +6
-0@@ -384,6 +384,12 @@struct smb_sb_info *server = req->rq_server;int result; +if (req->rq_bytes_sent ==0 ) { +WSET (req->rq_header, smb_tid, server->opt.tid); +WSET (req->rq_header, smb_pid,1 ); +WSET (req->rq_header, smb_uid, server->opt.server_uid); + } + result =smb_send_request (req);if (result <0 && result != -EAGAIN ) goto out;diff -urN -X exclude
linux-2.6.3-rc1-orig/fs/smbfs/smbiod.clinux-2.6.3-rc1-smbfs/fs/smbfs/smbiod.cfs/smbfs/smbiod.c +3
-0@@ -161,6 +161,9 @@while (head != &server->xmitq) { req = list_entry(head, struct smb_request, rq_queue); head = head->next; + +WSET (req->rq_header, smb_uid,-1 ); + req->rq_bytes_sent =0 ;if (req->rq_flags &SMB_REQ_NORETRY ) {VERBOSE ("aborting request %p on xmitq\n" , req); req->rq_errno = -EIO ;
Quoted 2 lines · Nico Schottelius
Works!
Will it be included into the main tree?
Yes, unless someone complains.
Also, I'm waiting for feedback on the smb_uid part (that I didn't tell you about) so I don't think it will be in 2.6.3.
/Urban