| Issue 42: | ntfs-3g doesn't respond to SIGTERM | |
| 2 people starred this issue and may be notified of changes. | Back to list |
What steps will reproduce the problem? reboot What is the expected output? What do you see instead? expected nothing, instead i see "ntfs-3g doesn't respond to SIGTERM" upon reboot/shutdown, causes a 10 - 15 secs lag. What version of the product are you using? On what operating system? MacFUSE 0.1.0_beta6-r3.dmg and NTFS-3G 20070116-r4.dmg from http://forum.insanelymac.com/index.php?showtopic=38920&hl=macfuse Please provide any additional information below. also my spotlight behaves badly / doesn't work / doesn't add new files into index. Don't know anything more specific. |
|
,
Jan 21, 2007
maybe I should have posted this to ntf3g project... |
|
,
Jan 21, 2007
OK, another problem, I can't create files bigger than 4GB on the NTFS partition. I think that has to do sth with fuse/ntfs-3g implementation on macosx rather than with ntfs-3g driver itself. Maybe of use would be to mention that if I delete ntfs.fs (i.e. the original one from apple) then ntfs-3g doesn't work. Is there any dependency (which also causes the 4GB limit)? |
|
,
Jan 22, 2007
Yes, please don't report issues here that are clearly unrelated to MacFUSE. As for the 4GB thing, some relevant things were changed in the newest release (MacFUSE Core 0.1.7). Please upgrade and see if the > 4GB stuff works fine. |
|
,
Jan 23, 2007
OK, installed Core 0.1.7 and NTFS-3G 20070116-r4.dmg (sudo rm'ed all files installed before and deleted receipts). I could create an 8 GB sparse image (which I couldn't before). Also tried to create a 7 GB toast image (from a DVD) - it tries to allocate space first and splits into parts (toast problem?). Then tried to create a 4.37 GB DVD-R .dmg image (full size read-write). It took insane amount of time, activity monitor showed constant 11.5 MB write speed and 18.5 GB (!!!) written in total when finished. Nevertheless the file is bigger than 4GB and it is there. I'm still getting the 0 KB size/free space thing. Creating a new folder/file in other application than finder (or vice versa) doesn't refresh the filesystem. I have to log out/in to see it or quit/relaunch the application in question for it to see the changes. Sorry for reporting things which are/could be more related to ntfs-3g rather than to the macfuse project itself - the thing is I don't have much knowledge about all this and I assume that people don't have these problems on linux (i.e. that they are macosx specific). So I'm just reporting problems which I noticed. Please understand that my interest (as well as interest of many others) is focused onto the so much needed ntfs write capability. Note: i didn't compile anything myself (don't know how to nor have the tools), I assume that NTFS-3G 20070116-r4.dmg is compatible with 0.1.7 (i.e. doesn't need recompile). |
|
,
Jan 23, 2007
MacFUSE Core 0.1.7 fixes the 0 KB size/free space issue by actively waiting for the file system to initialize before calling it "done". If the caller (ntfs-3g in this case) is using the ping_diskarb option, then the caller doesn't have to do anything, as MacFUSE will let Disk Arbitration know at the right time. If, however, the caller is *not* using ping_diskarb (which I think is the case with ntfs-3g here), then the caller has to wait itself before proceeding. 0.1.7 adds a system-wide notification that's sent to the distributed notification center in Mac OS X when the mount is truly complete. The caller should be an observer for this notification and wait. The reason you're still seeing the 0 KB thing is because Disk Arbitration is likely being pinged too early still. |
|
,
Jan 23, 2007
Though the problems described here aren't necessarily the fault of MacFUSE or NTFS-3G, one solution lies in fusefs.kext. The SIGTERM problem extends from the fact that ntfs-3g userspace daemon does not respond to SIGTERM when OS X tries to shut down. While other daemons might, NTFS-3G doesn't. What's being revealed here is a layering violation of sorts. NTFS-3G volumes are mounted in userspace, but refuse to die like a userspace program. Thus, to eliminate the error, FUSE filesystems should be unmounted when the shutdown procedure initiates. There are two ways to do this. One is by creating an NTFS-3G StartupItem that only responds to the "stop" request, and subsequently unmounts NTFS-3G filesystems. Alternatively, and likely more efficient, is having fusefs.kext register for shutdown notifications from IOKit, so that *all* FUSE filesystems can be unmounted prior to shutdown. Would it be possible to add unmount-on-shutdown (or restart) support to fusefs.kext to avoid seeing errors like this in other FUSE filesystems? |
|
,
Jan 24, 2007
Hopefully with the notifications in place, this can be handled on the ntfs-3g side.
Status: WontFix
|
|
|
|