ioFTPD v6.2.0 Beta Released

[oldhouse]

Originally Posted by oldhouse

09-23-2007 22:08:51 STOP: "PID=224"
09-23-2007 22:26:37 START: "PID=2200" "CmdLine="

It started only after I killed it in task manager, even if ioFTPD.exe was already active.

Scheduler update didn't have any effect.

no more suggestions for this problem?

[PSA9]

Yil i see that i am going to run into a problem and i want to know if this workaround will or will not work.

when i log into my FTP these are the only things that will ever be seen at the main

_INCOMPLETE
_SORTED
-=[REQUEST]=-
Applications
Books
Games
Movies
Music
Music Videos

3 750gb, 3 400gb, and 1 300gb will be used. a old method i used in the .VFS was like say
X:Movies /Movies
Y:Movies /Movies

this worked fine if downloading only but as u know uploading would only goto say the Y drive. this problem has been talked about before, but i never seen anyone talk of a solution to this or if its going to be fixed. so i would like to know if this workaround would work, or if u could add support to make it work.

X:/Movies /Movies/DVDR
Y:/Movies /Movies/HDDVD

while i do know its possible to do that, what i dont know is if say i was to upload something to the DVDR folder if it would send it to the X drive? if not can u add support for this?

MY Next ? is SSL, since i am running a 64 bit vista, would i have to make a 64 bit cert, or is SSL not dependent on 32 bit or 64 bit and would work same as vista 32 bit? either way can u direct me to where to read on how to make certs for Vista!
i read this in ioftpd.ini

Don't set min higher than 128 or non-Vista FTP clients not built with openSSL

with that being said, if i set to 256, will non-vista users using FlashFXP would or would not be able to connect? i dont want anyone connecting without flashFXP, and if their was a serious need, i could always exempt just those users from need SSL to login!

[Yil]

mr_F: I'll play around with it, but just want to make sure you're running under the same user account you deleted the directory with in explorer and that it was a local dir. Ditto for the source dirs. Unlocker/procexp coming up clean is good. I was thinking google desktop or some other indexing app was running around in the background trying to index stuff.

Oh! Just had an idea. Do you have show hidden files turned on AND Hide protected operating system files turned OFF? I'm betting there is a thumbs.db file or something that you can't see in the directory. ioFTPD sees that the directory isn't empty and bails. That's probably a bug. Assuming that is the only entry besides .ioFTPD* files it should treat the directory as empty and let it be deleted...


oldhouse: nope, got me. What does the service event log show? If you use "netstat -a" do you see the port in use? Does "netstat -ab" show the right executable and pid? Go grab tcpview.exe from Microsoft (it used to be a sysinternals app) for a much simpler interface than netstat to view the details... Think up as many things as you can to help define what is going on better.

[Yil]

PSA9: Great question!!! Provided /Movies exists as a real directory or as a virtually mounted directory in the VFS you can do what you want (I don't think you can get away without having one). Users will see /Movies having 2 subdirs DVDR and HDDVD and uploads will go to the correct drive.

X:\Movies /Movies/DVDR
Y:\Movies /Movies/HDDVD

You can ALSO use

X:\Movies /Movies
Y:\Movies /Movies

with directories x:\Movies\DVDR and y:\Movies\HDDVD and achieve the same thing. It's a lot trickier though if you later create a DVDR dir on Y outside of ioFTPD since it will start accepting uploads over X with interesting results for subdirs. I detailed some of these issues in the default.vfs file.

I currently have some cool new features partially working to really solve this problem though. I'm hoping for the 6.3 release to finish them up, but it's tricky code so I'm thinking of pulling it and waiting for 6.4. Here's the relevant entries from the ioftpd.ini file:

# Format for Min_Space_* entries is a number followed by at least one space
# and then a double quoted list of '|' separated, wildcard enabled paths to
# apply the limit to.  As special cases a single letter is presumed to be a
# drive letter and thus 'C' is equivalent to 'C:\*' and all entries ending
# in a '\' are interpretted as ending in a wildcard (i.e. '...\*'). 
#
# Example:
#   ... = 50 "C|D"   5000 "G:\ISOs\|G:\DVDRS\"    100 "*"
#   This would apply 50MB to anything on drives C and D, 5000 to anything
#   under the directories ISOs or DVDRS on drive G, and 100 to any other
#   drive in the system like E, F, as well as any other dirs on drive G.
 
# Required minimum free disk space (in MB) for accepting new uploaded files.
# Once this limit is reached further uploads are rejected just as if the
# drive was actually out of disk space.
;Min_Space_For_Upload = 20 "*"
 
# Required minimum free disk space (in MB) for approving the creation of
# new directories.  Once this limit is reached further directories are
# denied just as if the drive was actually out of disk space.
;Min_Space_For_Dir = 100 "*"
 
 
# The Span_Deny and Span_Approve rules work in tandem to automatically enable
# new directories in merged/raided virtual mounts (i.e. multiple paths mounted
# on the same virtual path so that their contents appear merged) to failover
# to one of the other paths if the currently resolved path becomes full and
# fails the Min_Space_For_Dir test.
#
# Algorithm:
#   Min_Space_For_Dir is defined and the test fails for the new dir's path
#   Span_Approve is defined and matches VFS path
#     [if not defined then reject the directory as spanning isn't enabled]
#   Span_Deny isn't defined or is and there is no match for the VFS path
#     [if defined and it DOES match the VFS path then the new directory is
#       presumed to be a multi-dir release and the directory is allowed as
#       a subdirectory or wherever the parent is].
#   Now search from the bottom up of the merged directories in the vfs file
#     to find a directory that passes it's own Min_Space_For_Dir test.
#     [if none found reject directory]
#   Any necessary parent directories are created on the new path
#   The MKD FTP_Pre-Command_Event is called if one is specified
#     [if the directory is rejected, any created parent directories are
#      silently removed if possible]
#   The directory is created normally from this point on
#
# This is a bit complex because care must be taken not to split certain
# directories across physical filesystems.  First off, since this logic only
# applies to new directories zipscripts shouldn't be effected as files
# in any particular directory will never be separated from each other.
# The problem lies in multiple disk releases.  Spanning /ISOs/08xx so that
# some immediate subdirs are on drive D and other on drive E is wonderful!
# However splitting /ISOs/08xx/Linux-v2.6 so that CD1 and CD2 is on D and
# cd 3,4,5,6 are on E probably isn't a good thing...
#
# When done right the end user will never notice any difference and they
# can just go ahead and keep uploading as you fill up drive after drive!
#
# NOTE: all VFS directory paths (hence everything we are talking about here)
#       always end in a '/'. Space around the '|' is ignored so you can format
#       with some hope of readability.
#
# Example1:
#  Span_Approve = /ISOs/*
#  Span_Deny    = /ISOs/*/*/*
#
# Allow just the ISOs directory to automatically span (even if the .vfs file
# lists other virtual directories with multiple mount points), but try to
# catch any any directory more than 3 levels deep so that
# /ISOs/08xx/Linux-v2.6/ is OK, but /ISOs/08xx/Linux-v2.6/CD1/ is not and
# thus CD1 won't separate from it's parent.
#
# Example2:
#  Span_Approve = *
#  Span_Deny    = */cd1/* | */cd2/* 
#
# Allow spanning whenever possible in the .vfs, but allow any path with any
# of the listed subdirectories to stay with it's parent.
 
# A '|' separated list of wildcard enabled VFS paths that are OK to span.
;Span_Approve = *
 
# A '|' separated list of wildcard enabled VFS paths that shouldn't be spanned.
;Span_Deny    = */cd1/* | */cd2/* | */cd3/*

As you can see to do this properly while not breaking any zipscripts took a little work and it's still subject to change... The biggest problem I have is what happens when users start deleting old stuff off the first totally filled up drive. The drive then has free space and will start accepting new directories and uploads again but this would tend to split up stuff that I'd probably like to see going next to whatever was just uploaded because it's easier to find outside of ioFTPD that way. So I need to introduce some sort of memory or drive affinity or something that I haven't figured out yet. It's not that it wouldn't work now, but it doesn't work the way I want it to yet.

This functionality is something that no script could currently perform because there simply isn't a way to redirect the target of a file / directory command. You can approve the action, but not modify it. I'm tempted to make a way to do this but it would have to be a different section of the .ini file or something because it would break existing scripts...

If you can think of a better way to do this or see a problem I missed do speak up now

[b0rz_]

Source code are avaible ? Just for fun i mean .... but, bc is a GPL2 project, source would be fine to have

[Flow]

Yil!, *amen*

[Yil]

bOrz: There's a link to the 6.1 source in the 6.1 thread. 6.3 will have a new source release as well.

[Zer0Racer]

Yil, I just stumbled onto something strange. A dir suddenly gives me "550 No such file or directory" when I try to enter it or issue any command on it, like chmod or whatever. Though the dir exists both in regular dir listing and raw dir list.

Io has been up for a little more than a week and, to my knowledge, nothing strange has happened lately. Worked just fine yesterday when I uploaded stuff to that particular dir.

Any ideas?

/ZR

P.S.
Allowed_Recursive = !*
No_SubDir_Sizing = False
D.S.

[Yil]

Hmm... If this is a simple VFS mount then I don't know, just restart it I guess. If it's a raided/merged directory then there are situations where that can occur but somehow I think I remember you saying your setup was pretty straightforward.

[Zer0Racer]

You remember correctly, no raided dirs or anything like that here. Yes, a restart would probably straight it out, but I don't want to restart if this is some kind of bug I guess it would be pretty hard to find it.

/ZR

EDIT: I tried Unlocker without success. Then I renamed the dir locally (just added an underscore last) and after that I could CWD into it and list etc. and io recached it. Btw, it's a subdir to a mount in io. All dirs are physical though, no virtual mount points or anything like that. Then I renamed it back to its original name and everything was back to normal. One thing I forgot to mention is that I do have a symlink (the new kind) in ftp root to that particular dir. It has of course always worked (but didn't when I couldn't cwd into the dir). It started to work again after the renaming back and forth.

So that's a work-around for now.

[PSA9]

Zer0Racer i have seen this problem before i did my massive upgrade i just never said anything as i was going to wait till i did my clean install. well i notice a similar problem just as u speak. i would say da FTP been on for almost a week, i did some heavy transfers on it to test, and at most 3 users were on but not doing heavy things.

say with ioSFV the Incomplete symlink dir, i having that problem when i try and enter them from USER 0, but USER 1 i could enter. USER 1 is the owner of the files so to further test this, i tried with USER 2, 3 and 4 which where all succesful. so i then again tried USER 0 with no luck. USER 3 is in same group as user 0 not that, that matters but all other users are in own group. i am not home to do a restart or anything else of such. i do know that i did add USER 5 as the first user that i didn't assign a group to which put him in the NoGroup 15 min before i noticed this. not saying this is a cause but just for incase.

another thing i am noticing and even after a rehash of da config.
right now as i speak i am offshore at work, loged in as USER 0 and USER 1 to say FXP and look at some things. USER 0 has its own .vfs while all other users have default. something else i should mention for that above problem. but their is only 2 users loged into my ftp, USER 0 and USER 1 on the same flashfxp. i had restarted the pc i am on and opened up one flash fxp and no one else is connected. this means that only 2 users are loged in but it shows that 5 users is loged in. 4 shows that USER 1 is loged in. say 1-2 hours ago before i rebooted this pc, i had 2 instances of flashfXP open and i had connected then, but its been like way too long for it to show that it is still loged in.

PLEASE CHECK into these problems. ooh something maybe i should mention the USER that is being showed as log in but is not logged in, is da only idle exempt user!!!

[Yil]

Multiple processors/cores. Poking around and trying to get my local server to crash has been unsuccessful but a quad processor systems seems to hang pretty often. I don't know why yet though, but it got me thinking today. With the advent of real multiprocessor systems via the explosion of dual/quad core systems instead of the fake Intel hyperthreading stuff it's entirely possible that a number of latent bugs are coming to the surface that just didn't get exposed too often on single core systems. Assuming a rare race condition it has to be more likely that this would be an issue in a dual core or dual processor system than any single core machine. If you have had repeated problems with crashes/hanging/etc please let me know what type of setup you have. Perhaps it will become obvious that multi-cores crash way more often than single cores and that will give a hint about why I have so much trouble reproducing some things locally.

[Flow]

Hi Yil, wanna release a update pack before poking more into cpu cores?

[mr_F]

Originally Posted by Yil

mr_F: I'll play around with it, but just want to make sure you're running under the same user account you deleted the directory with in explorer and that it was a local dir. Ditto for the source dirs. Unlocker/procexp coming up clean is good. I was thinking google desktop or some other indexing app was running around in the background trying to index stuff.

Oh! Just had an idea. Do you have show hidden files turned on AND Hide protected operating system files turned OFF? I'm betting there is a thumbs.db file or something that you can't see in the directory. ioFTPD sees that the directory isn't empty and bails. That's probably a bug. Assuming that is the only entry besides .ioFTPD* files it should treat the directory as empty and let it be deleted...

yil:
sorry it took me so long, but yes it was the same user that is running ioftpd that i deleted the directory under in explorer -- and the directory is indeed a local directory.

furthermore i created a directory that after access contained a thumb.db, and alternatively, a folder with a desktop.ini. you are correct, ioftpd could not delete either of these under both circumstances: that hidden system files were visible and invisible.

after further testing, it appears that any file that is not the .ioftpd file which contains the 'hidden' attribute is undeletable by ioftpd, even if visible in the dir listing:

The following file was visible in dir listing, but contained the 'hidden' attribute:
[20:21:12] [L] DELE 35071068.JPG
[20:21:12] [L] 550 35071068.JPG: No such file or directory.

after removing the attribute with SSH:
[20:22:07] [L] DELE 35071068.JPG
[20:22:07] [L] 250 DELE command successful.

it might be OS specific to Vista, not sure. i hope my findings help you

-mr_F

[rolan]

hi, Yil
coz always crash, so i not setup any scripts, but remain crash.

Unhandled exception 3221225477 at address 0x7c94bd02 (0x00000000)
0x00000000 0x00000003

and user ioFTPD v6.2.1