Anaconda-devel March 2008

anaconda-devel@lists.stg.fedoraproject.org

24 participants
28 discussions

[ PATCH ] RFC: Search and load drivers automatically from usb-storage media
by Sandeep_K_Shandilya＠Dell.com 11 Apr '08

11 Apr '08

Hello All I have moved this discussion from bugzilla to the mailing list Currently customers have to use driver disks or usb-storage to install to newer hardware (NICS and disk controllers) for which the OS does not have support. The customer has to manually prepare driver disks or give boot time command line options to locate these driver disk images. This method could be improved. Here are some of the questions Jeremy had with regard to this patch. Please read inline <sandeep> A few comments from a quick look: 1) Instead of posting patches in private bugs, it's far better to send them to anaconda-devel-list so they can actually be looked at and commented on by the community, not just some private Dell/Red Hat interaction, that is what I have done. 2) Automatic mounting of partitions with no way to opt-out can lead to problems over time <sandeep> Do you have any specific areas in the anaconda loader that you are referrring to? we will anyway unmount the partition after copying the driver updates to ramfs. 3) The naming is terrible. What's OEM specific about this functionality? <Sandeep> Well, nothing OEM about it except the OEM's would like to install the drivers that they have tested and qualified with NEW hardware. 4) What's better about doing this automatically as opposed to actually having the user specify that "yes, I have a driver disk" <Sandeep> Think about it this way, Suppose the drivers with new hardware support are placed in a utility partition (OEM will prepare this on the system) on an embedded USB storage device, or thinking a little bit more into the future the driver may reside on the internet Dell is already pursuing this and Dell servers to be released in future may have embedded usb-storage with drivers, diagnostics etc... regards, Sandeep

5 30

on creating driver disks
by John Summerfield 03 Apr '08

03 Apr '08

I had in mind the idea of doing a network install of CentOS5. The primary impediment is that the target machine's going to be wireless just as soon as I stick my wireless card in it. Support for the card's not an issue, the prism54 driver in Ubuntu Warty's been working fine for years. What I think I would like is a driver disk. Not really knowing how to create one, I asked Google. Google is full of helpful advice "download dd.img and write it to a floppy" that misses the mark, and I found Doug Ledford's kit and comments which is a bit nearer the mark, but not really. I've read through it and got a bit of an idea, but I think it's not going to work. I need the driver (and I'm supposing I have a suitable one already built, but if not then I _can_ address that), and I need to load firmware for the wireless card. Now I'm sure I can open up the initrd and add the necessary magic, but that doesn't seem the One True Way. In the interests of getting this working before me cheese gets back, I'm going to work around the issue, but it does raise some questions" 1. Is Doug's documentation the best there is? 2. Assuming I popped the right prism53.ko and appropriate files (including firmware) on to a floppy disk in the proper layout, would it actually work? 3. Where is the layout of the disk described? 4. Where is the content of the support files described? -- CU John -- Cheers John -- spambait 1aaaaaaa(a)coco.merseine.nu Z1aaaaaaa(a)coco.merseine.nu -- Advice http://webfoot.com/advice/email.top.php http://www.catb.org/~esr/faqs/smart-questions.html http://support.microsoft.com/kb/555375 You cannot reply off-list:-)

3 3

[PATCH] simplify ext4 install boot option
by Eric Sandeen 31 Mar '08

31 Mar '08

I think just "ext4" is enough at this point... migration seems totally busted; the root fs and everything in its fstab is mounted before the migration routine runs, so it doesn't get its fs type changed until the *next* mount, which for this migration path pretty well misses the window of opportunity. :( So I figure maybe "ext4migrate" to enable it, but it's not really going to do what you might hope... Thanks, -Eric Index: anaconda-11.4.0.64/fsset.py =================================================================== --- anaconda-11.4.0.64.orig/fsset.py +++ anaconda-11.4.0.64/fsset.py @@ -722,7 +722,7 @@ class ext3FileSystem(extFileSystem): self.name = "ext3" self.extraFormatArgs = [ "-j" ] self.partedFileSystemType = parted.file_system_type_get("ext3") - if flags.cmdline.has_key("iamanext4developer"): + if flags.cmdline.has_key("ext4migrate"): self.migratetofs = ['ext4dev'] def formatDevice(self, entry, progress, chroot='/'): @@ -758,7 +758,7 @@ class ext4FileSystem(extFileSystem): self.extraFormatArgs = [ "-j", "-I", "256", "-E", "test_fs" ] # this is way way experimental at present... - if flags.cmdline.has_key("iamanext4developer"): + if flags.cmdline.has_key("ext4") or flags.cmdline.has_key("ext4migrate"): self.supported = -1 else: self.supported = 0

2 2

[PATCH] Update pkgorder group listings to match current Fedora defaults.
by Jesse Keating 31 Mar '08

31 Mar '08

This brings default installs back to just first three CDs, 2 CDs if Office and Productivity is deselected, and one CD if everything is deselected. --- scripts/pkgorder | 13 +++++++------ 1 files changed, 7 insertions(+), 6 deletions(-) diff --git a/scripts/pkgorder b/scripts/pkgorder index 4cbca0d..f72249c 100755 --- a/scripts/pkgorder +++ b/scripts/pkgorder @@ -182,15 +182,15 @@ if __name__ == "__main__": addGroups(ds, ["base-x", "dial-up", "graphical-internet", "editors", - "graphics", "gnome-desktop", "sound-and-video", - "printing", "fonts", "hardware-support"]) + "gnome-desktop", "sound-and-video", "printing", + "fonts", "hardware-support", "admin-tools", + "java", "legacy-fonts"]) - addGroups(ds, ["office", "engineering-and-scientific", - "authoring-and-publishing", "games"]) + addGroups(ds, ["office", "games", "graphics", "authoring-and-publishing"]) addGroups(ds, ["web-server", "ftp-server", "sql-server", "mysql", "server-cfg", "dns-server", - "smb-server", "admin-tools"]) + "smb-server"]) addGroups(ds, ["kde-desktop", "development-tools", "development-libs", "gnome-software-development", "eclipse", @@ -198,7 +198,8 @@ if __name__ == "__main__": "java-development", "kde-software-development", "mail-server", "network-server", "legacy-network-server"]) - addGroups(ds, ["news-server", "legacy-software-development"]) + addGroups(ds, ["news-server", "legacy-software-development", + "engineering-and-scientific"]) #Everthing else but kernels for po in ds.pkgSack.returnPackages(): -- 1.5.4.5

2 1

Why can't the hard disk partition use the same logic as NFS and URL installs?
by G.Wolfe Woodbury 30 Mar '08

30 Mar '08

The "local disk" installation option is limited to using ISO images for installation, while NFS and URL (http and ftp) installs will use *either* ISOs or repository trees. The question is why? To me it appears that this is a very limiting situation, and that the local partition should also be able to contain *either* form of package sources. Is is just that the local partition option is buried so deep in the heart of history that it can't be changed easily? I would think that the developers would simply find a way to specify the source of packages once and for all, and then let the rest of the install use a uniform interface. After all, the current method is to loop mount ISOs and then access packages as if they were repos. Again, why not use a local partition the same way. Of course, it would not be prudent to read and write to the sources partition at the same time, but I suspect that the experienced users could live with a restriction that the source not be on a partition that is going to be the target of an install. Allowing "local partition" installs to use either form would enable something like using a USB storage unit to be the source for multiple installs. (I haven't the skills to actually modify the code, or else I would submit patches. I'm an administrator and tester, using the installer a lot, no longer a developer or coder.) -- G.Wolfe Woodbury

2 1

kickstart help
by Hoy, Mark 28 Mar '08

28 Mar '08

I'm attempting to build a customized kickstart that copies the entire CD to the hard drive before finishing. While reading the RH guide on anaconda I've not found what I'm looking for. A part of the kickstart file that runs while the sources are still at /mnt/sources. Ideally if there was a "%post --mntsource" section existed I'd be a happy guy. Instead I have some code in "%post --nochroot" section that figures out the CD rom location and re-mounts it, (/dev/hdc or something similar). Difficulties encountered -real hardware and vmware have different mount points for the cdrom. -what happens when run over a network? Using CentOS 5.1 with anaconda-11.1.2.87-1.el5.centos.i386.rpm anaconda-runtime-11.1.2.87-1.el5.centos.i386.rpm : Mark

1 0

[elliot@rpath.com: anaconda: when dhcp is selected ensure that bootproto is set to dhcp (RPL-2301)]
by Elliot Peele 27 Mar '08

27 Mar '08

If Anaconda was started without network configured the netdev object does not have a bootproto set. The patch ensures that bootproto is set for the dhcp case. In the static case the ui requires the user to enter ip information. Elliot tag: anaconda-2008032601 user: Elliot Peele <http://issues.rpath.com/> files: iw/netconfig_dialog.py when dhcp is selected ensure that bootproto is set to dhcp (RPL-2301) diff -r 3e60e4b058e2 -r bcba0cd916c1 iw/netconfig_dialog.py --- a/iw/netconfig_dialog.py Wed Mar 26 16:02:08 2008 -0400 +++ b/iw/netconfig_dialog.py Wed Mar 26 16:03:44 2008 -0400 @@ -165,6 +165,7 @@ # FIXME: need to do input validation if self.xml.get_widget("dhcpCheckbutton").get_active(): + netdev.set(('bootproto', 'dhcp')) self.window.hide() w = gui.WaitWindow(_("Dynamic IP"), _("Sending request for IP information " -- Elliot Peele rPath, Inc. elliot(a)rpath.com

2 1

[elliot@rpath.com: anaconda: Assume there is link on devices that don't support the required ioctls]
by Elliot Peele 27 Mar '08

27 Mar '08

Anaconda currently does not detect link for drivers that do not support the IOCTLs. I found that in QEMU when trying to test iSCSI Anaconda tried to setup networking even if the network was configured. This patch assumes that if there is link on devices that don't support the required ioctls. Elliot user: Elliot Peele <http://issues.rpath.com/> files: isys/linkdetect.c Assume there is link on devices that don't support the required ioctls (RPL-2301) diff -r 925b8bd510d7 -r 3e60e4b058e2 isys/linkdetect.c --- a/isys/linkdetect.c Mon Mar 10 12:29:50 2008 -0400 +++ b/isys/linkdetect.c Wed Mar 26 16:02:08 2008 -0400 @@ -73,6 +73,8 @@ int i, mii_val[32]; if (ioctl(sock, SIOCGMIIPHY, &ifr) < 0) { + if (errno == EOPNOTSUPP) + return -2; if (errno != ENODEV) #ifdef STANDALONE fprintf(stderr, "SIOCGMIIPHY on '%s' failed: %s\n", @@ -113,6 +115,9 @@ #ifdef STANDALONE fprintf(stderr, "Cannot get link status (%d): %s\n", errno, strerror(errno)); #endif + } else { + // ethtool not supported by driver + return -2; } return -1; @@ -122,6 +127,7 @@ int get_link_status(char * devname) { int sock, rc; + int link_status_supported = 1; if ((sock = socket(AF_INET, SOCK_DGRAM, 0)) < 0) { #ifdef STANDALONE @@ -145,13 +151,15 @@ if (rc == 1) { close(sock); return 1; + } else if (rc == -2) { + link_status_supported = 0; } rc = get_mii_link_status(sock); #ifdef STANDALONE printf("MII link status of %s is: %d\n", devname, rc); #endif - if (rc == 1) { + if (rc == 1 || (rc == -2 && link_status_supported == 0)) { close(sock); return 1; } -- Elliot Peele rPath, Inc. elliot(a)rpath.com

4 4

further mouse extermination
by Bill Nottingham 27 Mar '08

27 Mar '08

Coming out of some of the rhpxl de-kudzu discussions, this patchset gets rid of even more of the pesky mouse infestation. Bill

2 3

Add support for preexisting encrypted devices
by David Lehman 27 Mar '08

27 Mar '08

This adds support for partitioning with preexisting encrypted devices. The basic idea is that we scan for encrypted devices before reading the initial partition layout in partitions.setFromDisk. We prompt for passphrases for encrypted devic es we find. Those devices for which we obtain a valid passphrase get the encrypt ion data added to the RequestSpec. The information is then used by the partition ing interface so users can create filesytems on these preexisting encrypted devi ces. They can also remove the encryption from said devices, provided they are wi lling to reformat the filesystem. It is also possible to do one edit of a partit ion to delete the LUKS header and then, on a subsequent edit, create a new LUKS header on the same device. The main thing that needs work IMO is some dependency resolution to provide orde ring for device access. Currently we just loop through opening all encrypted dev ices twice -- once before activating RAID and LVM, and once afterward. Ideally, we could see which devices need RAID and/or LVM to be active before we can try t o access them. The hammer seems to do the job in the meantime. Comments are welcome, of course.

2 6

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Anaconda-devel March 2008