Description of problem: This message quoted in summary does not want to go away already for some time on x86_64. On i386 does not seem to be present. A corresponding fragment of startup messages looks like follows: .... Starting udev: [ OK ] Initializing hardware... storage network audio done [ OK ] Loading default keymap (emacs2): [ OK ] Setting hostname localhost.localdomain: [ OK ] Setting up Logical Volume Management: File descriptor 53 left open No volume groups found .... ???? Version-Release number of selected component (if applicable): initscripts-8.15-1
Assigning to lvm2 for the moment; that's coming from the LVM tools.
Can you debug this by adding a line *immediately* before the lvm command in the initscripts to run '/usr/sbin/lsof' (no args) and capture its complete output to a file and attach it to this bug report?
And say whether you've built a special initrd, and if so, what options you used.
And which version of mkinitrd package you have
Created attachment 119556 [details] lsof output a moment before lvm commands are called /dev/sda5 which shows up in an attached output with "53r" is actually my swap partition. All initrd images were build by %post script in an installation process of new kernels. The current version of mkinitrd is mkinitrd-5.0.3-1. and the image used in testing was build with this one, but "File descriptor 53 left open" was showing up with earlier versions of mkinitrd as well and I was just forgetting to report it until mkinitrd really started to misbehave (see bug #169712). OTOH I have one initrd image build with mkinitrd-5.0.0-1 and, when I started careful checks, when booting with this one I do not see that "... left open" and there is no /dev/sda5 anywhere in an lsof output. The message surely was present too when various earlier versions were used and it is back if I am booting a kernel which has initrd made with a version 5.0.2-1 of mkinitrd (with /dev/sda5 splattered all over lsof output).
The relevant lsof line is: init 1 root 53r BLK 8,5 935 /dev/sda5 which suggests 'nash' in the initrd is forgetting to close this before starting 'init'.
Is the swap activated correctly after booting? (cat /proc/swaps)
Swap looks all right after a boot; both in /proc/swaps and in an output of 'free'. I observe that on x86_64 but this may be really a hardware difference. On that machine there is a SATA drive, which shows as /dev/sda, and the other box has a plain old IDE.
With the current version of startup (initrd and initscripts) I do not see, at this moment, this complaint. OTOH checking with lsof shows that init still holds open a block device of my swap.
*** Bug 165912 has been marked as a duplicate of this bug. ***
I am seeing looks of selinux messages becuase of this bug. To see the messages goto https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=165912 which is closed since it is a duplicate of this bug. I am running rawhide on i686 updated until yesterday (todays updates dont install due to dependency issues). I have initscripts-8.17-1 and mkinitrd-5.0.8-1 installed. I am using kernel-2.6.14-1635-1. Thanks, Jason
When I boot I get File descriptor 42 left open. Is this different than File descriptor 53?
> When I boot I get File descriptor 42 left open. Is this different than File descriptor 53? Most likely not. A particular number depends on your hardware and possibly on a kernel architecture too. You may check with lsof if this is a reference to your swap device. If this is really not then you are seeing something diffrent.
*** Bug 171659 has been marked as a duplicate of this bug. ***