Anaconda-devel June 2014

anaconda-devel@lists.stg.fedoraproject.org

9 participants
12 discussions

pdb.set_trace() with liveinst
by Gene Czarcinski 20 Sep '14

20 Sep '14

Is it possible to use pdb.set_trace() when liveinst is used to run anaconda? I tried and, if it is suppose to work, there is a bug because it blows up and exits when it hits the first pdb.set_trace() How do you debug when you have the liveinst environment? Gene

3 5

[PATCH 0/9] v2.3: btrfs and LVMlv support in grubby
by Gene Czarcinski 29 Jun '14

29 Jun '14

The actual startup for Fedora 21 is rapidly approaching and I would like to see grubby updated in rawhide sooner rather than later so that the full testing process can be used to reveal any "undesirable features." The version of anaconda installer in rawhide has been updated to support /boot on both btrfs and an LVMlv. The primary objective of this update is to add support for /boot on btrfs. However, it also includes some related updates with respect to LVMlv support and the basic operation of grubby. Rather than having one giant update, this has been broken up into 8 patch files which (hopefully) will be easier to understand. 1. Add code to limit the scan of a grub2 configuration file for a valid template to that to the main menu section. 2. Add support for /boot on btrfs subvol or <dir> under rootrf on btrfs subvol. 3. Add compile-time enabled debugging for btrfs support 4. Add tests for btrfs support 5. Add new run-time option to disable output from compile-time enabling of DEBUG. 6. Enable DEBUG (for now a good idea but can be disabled once there is confidence in new updates) ... perhaps DEBUG should always be enabled and a new runtime parameter enables output. For now it is what it is. 7. Similar to the regression tests for BTRFS support, add regression tests for /boot on LVMlv support. 8. The test files to support #7. 9. When an rpm is built under mock-chroot, DEBUG is enabled, andthe regression tests are run, some btrfs tests will fail. The problem is that findBootPrefi() does not work correctly under mock. Gene Czarcinski (9): v2 add code to validate a grub2 entry v2.2 add support for btrfs when grub2 is bootloader v2.1 add compile-time enabled debugging code for btrfs v2.1 add tests for btrfs support add disable-debug runtime option Enable DEBUG regression test to verify that boot on LVMlv supported files for LVMlv regressions tests bugfix findBootPrefix so it works under mock grubby.c | 319 +++++++++++++++++++++++++++++++++++++++++++---- test.sh | 78 +++++++++++- test/grub2.15 | 126 +++++++++++++++++++ test/grub2.16 | 140 +++++++++++++++++++++ test/grub2.17 | 128 +++++++++++++++++++ test/grub2.18 | 143 +++++++++++++++++++++ test/grub2.19 | 126 +++++++++++++++++++ test/grub2.20 | 128 +++++++++++++++++++ test/grub2.21 | 128 +++++++++++++++++++ test/grub2.22 | 143 +++++++++++++++++++++ test/grub2.23 | 128 +++++++++++++++++++ test/grub2.24 | 143 +++++++++++++++++++++ test/grub2.25 | 128 +++++++++++++++++++ test/grub2.26 | 128 +++++++++++++++++++ test/results/add/g2-1.15 | 140 +++++++++++++++++++++ test/results/add/g2-1.16 | 141 +++++++++++++++++++++ test/results/add/g2-1.17 | 143 +++++++++++++++++++++ test/results/add/g2-1.18 | 144 +++++++++++++++++++++ test/results/add/g2-1.19 | 141 +++++++++++++++++++++ test/results/add/g2-1.20 | 144 +++++++++++++++++++++ test/results/add/g2-1.21 | 143 +++++++++++++++++++++ test/results/add/g2-1.22 | 144 +++++++++++++++++++++ test/results/add/g2-1.23 | 143 +++++++++++++++++++++ test/results/add/g2-1.24 | 144 +++++++++++++++++++++ test/results/add/g2-1.25 | 144 +++++++++++++++++++++ test/results/add/g2-1.26 | 144 +++++++++++++++++++++ test/results/debug/g2.1 | 2 +- 27 files changed, 3675 insertions(+), 28 deletions(-) create mode 100644 test/grub2.15 create mode 100644 test/grub2.16 create mode 100644 test/grub2.17 create mode 100644 test/grub2.18 create mode 100644 test/grub2.19 create mode 100644 test/grub2.20 create mode 100644 test/grub2.21 create mode 100644 test/grub2.22 create mode 100644 test/grub2.23 create mode 100644 test/grub2.24 create mode 100644 test/grub2.25 create mode 100644 test/grub2.26 create mode 100644 test/results/add/g2-1.15 create mode 100644 test/results/add/g2-1.16 create mode 100644 test/results/add/g2-1.17 create mode 100644 test/results/add/g2-1.18 create mode 100644 test/results/add/g2-1.19 create mode 100644 test/results/add/g2-1.20 create mode 100644 test/results/add/g2-1.21 create mode 100644 test/results/add/g2-1.22 create mode 100644 test/results/add/g2-1.23 create mode 100644 test/results/add/g2-1.24 create mode 100644 test/results/add/g2-1.25 create mode 100644 test/results/add/g2-1.26 -- 1.9.3

1 9

[PATCH 0/8] v2.2: btrfs and LVMlv support in grubby
by Gene Czarcinski 29 Jun '14

29 Jun '14

The actual startup for Fedora 21 is rapidly approaching and I would like to see grubby updated in rawhide sooner rather than later so that the full testing process can be used to reveal any "undesirable features." The version of anaconda installer in rawhide has been updated to support /boot on both btrfs and an LVMlv. The primary objective of this update is to add support for /boot on btrfs. However, it also includes some related updates with respect to LVMlv support and the basic operation of grubby. Rather than having one giant update, this has been broken up into 8 patch files which (hopefully) will be easier to understand. 1. Add code to limit the scan of a grub2 configuration file for a valid template to that to the main menu section. 2. Add support for /boot on btrfs subvol or <dir> under rootrf on btrfs subvol. 3. Add compile-time enabled debugging for btrfs support 4. Add tests for btrfs support 5. Add new run-time option to disable output from compile-time enabling of DEBUG. 6. Enable DEBUG (for now a good idea but can be disabled once there is confidence in new updates) ... perhaps DEBUG should always be enabled and a new runtime parameter enables output. For now it is what it is. 7. Similar to the regression tests for BTRFS support, add regression tests for /boot on LVMlv support. 8. The test files to support #7. Gene Czarcinski (8): v2 add code to validate a grub2 entry v2.2 add support for btrfs when grub2 is bootloader v2.1 add compile-time enabled debugging code for btrfs v2.1 add tests for btrfs support add disable-debug runtime option Enable DEBUG regression test to verify that boot on LVMlv supported files for LVMlv regressions tests grubby.c | 312 +++++++++++++++++++++++++++++++++++++++++++---- test.sh | 78 +++++++++++- test/grub2.15 | 126 +++++++++++++++++++ test/grub2.16 | 140 +++++++++++++++++++++ test/grub2.17 | 128 +++++++++++++++++++ test/grub2.18 | 143 ++++++++++++++++++++++ test/grub2.19 | 126 +++++++++++++++++++ test/grub2.20 | 128 +++++++++++++++++++ test/grub2.21 | 128 +++++++++++++++++++ test/grub2.22 | 143 ++++++++++++++++++++++ test/grub2.23 | 128 +++++++++++++++++++ test/grub2.24 | 143 ++++++++++++++++++++++ test/grub2.25 | 128 +++++++++++++++++++ test/grub2.26 | 128 +++++++++++++++++++ test/results/add/g2-1.15 | 140 +++++++++++++++++++++ test/results/add/g2-1.16 | 141 +++++++++++++++++++++ test/results/add/g2-1.17 | 143 ++++++++++++++++++++++ test/results/add/g2-1.18 | 144 ++++++++++++++++++++++ test/results/add/g2-1.19 | 141 +++++++++++++++++++++ test/results/add/g2-1.20 | 144 ++++++++++++++++++++++ test/results/add/g2-1.21 | 143 ++++++++++++++++++++++ test/results/add/g2-1.22 | 144 ++++++++++++++++++++++ test/results/add/g2-1.23 | 143 ++++++++++++++++++++++ test/results/add/g2-1.24 | 144 ++++++++++++++++++++++ test/results/add/g2-1.25 | 144 ++++++++++++++++++++++ test/results/add/g2-1.26 | 144 ++++++++++++++++++++++ test/results/debug/g2.1 | 2 +- 27 files changed, 3669 insertions(+), 27 deletions(-) create mode 100644 test/grub2.15 create mode 100644 test/grub2.16 create mode 100644 test/grub2.17 create mode 100644 test/grub2.18 create mode 100644 test/grub2.19 create mode 100644 test/grub2.20 create mode 100644 test/grub2.21 create mode 100644 test/grub2.22 create mode 100644 test/grub2.23 create mode 100644 test/grub2.24 create mode 100644 test/grub2.25 create mode 100644 test/grub2.26 create mode 100644 test/results/add/g2-1.15 create mode 100644 test/results/add/g2-1.16 create mode 100644 test/results/add/g2-1.17 create mode 100644 test/results/add/g2-1.18 create mode 100644 test/results/add/g2-1.19 create mode 100644 test/results/add/g2-1.20 create mode 100644 test/results/add/g2-1.21 create mode 100644 test/results/add/g2-1.22 create mode 100644 test/results/add/g2-1.23 create mode 100644 test/results/add/g2-1.24 create mode 100644 test/results/add/g2-1.25 create mode 100644 test/results/add/g2-1.26 -- 1.9.3

1 9

grub2, 30_os-prober, os-prober: A Proposal
by Gene Czarcinski 23 Jun '14

23 Jun '14

While not specifically anaconda released, this mailing list was suggested to have a discussion on a proposal I would like to make. While I am writing this message, another individual has also been involved: Heydayat Vatankhah <hedayatv(a)gmail.com> who happens to be the Fedora package maintainer for the os-prober package. Currently, grub2's 30_os-prober and the os-prober package as implemented in rawhide are seriously broken. They simply do not work and an OS newly installed by anaconda based on the current rawhide will be unable to boot previously installed OSs. The BZ reports and attached patches to fix the problem are here: https://bugzilla.redhat.com/show_bug.cgi?id=1108296 https://bugzilla.redhat.com/show_bug.cgi?id=1108344 It takes both updates to fix things. Either one by themselves will not break things any worse than they are. This is not the first time that Heydayat and I have had problems with os-prober and we have sent the fixes upstream but they have been ignored. For example: https://bugzilla.redhat.com/show_bug.cgi?id=888341 and note the response from upstream debian. Anyway, Heyayat has previously mentioned forking os-prober and I believe that the time has com to do just that. 1. Create a new package os-prober-ng which takes the current os-prober-1.58-7 in rawhide plus my new patch plus a couple of the patches in the debian os-prober repository and use that as the basis for os-prober-ng-0.1. Naturally, lots of files will be renamed so that they are different from the files in os-prober. The intent is not to exactly replace os-prober but simply offer what I believe will be a better product. Any better suggestions for a different name are welcome. 2. The new os-prober-ng will be an information gathering tool intended to help configure a bootloader. For the most part, the current "os-prober" program looks to be fairly complete as is (except for a bug I am working on). However, it runs a number of tests for obscure operating systems which lengthen the time that it takes to run grub2-mkconfig. This includes qnx, lsb, hurd, minux, and Haiku. I suggest these be dropped or, at the very least, put into a separate binary rpm for optional installation. 3. An integral part of the current os-prober is the linux-boot-prober which gathers the information needed to create all those wonderful menuentry items by 30_os-prober. I have an alternative to propose (see below) and therefor intend to make this another optional binary rpm. 4. A significant new program will use the path to a systems rootfs and return the path to the grub.cfg for that system instance. This info will be used to "chain load" grub2 configuration files. 5. It is interesting that extlinux (syslinux) also has the capability to chain-load configuration files. Therefore, another optional program will return the information needed to create chain-loading extlinux configuration files. 6. Thanks to some prodding and information provided by Chris Murphy, I believe I have figured out a way to chain-load UEFI configuration files so that you will be able to have multi-boot fedora installations on a UEFI system. Right now this is based on some hand-crafted grub2 files but it does appear to work and I have a qemu-kvm-ovmf UEFI system with both Fedora 20 and Fedora 21 installed on a single disk (naturally it is BTRFS partitioned ;) ). Both boot and run in secure-boot mode. Beside os-prober-ng, I am proposing a new package which I call 'multi-boot-configtool". This package will use the outputs of os-prober-ng to create grub2 and (later) extlinux configurations capable of booting previously installed systems. 1. The software will use /etc/grub.d/* files to support most of the automation. However, there will be the capability to run some of it independent of grub2 and grub2-mkconfig. 2. The package will have a /etc/default/multi-boot-configtool file to store optional runtime variables. 3. The package will support creating configurations to "chainloader +1" MBR, BIOS Boot partitions and EFI system partitions. 4. There will be three types of (optional) configuration files: a) custom.cfg which is in the same directory as grub.cfg and is invoked dynamically by /etc/grub.d/41_custom, b) inline in a manner similar to the way 30_os-prober does, and c) into a file similar to /etc/grub.d/40_custom. 5. There will be a 30_multi-boot which will be a somewhat re-written 30_os-prober. I am also proposing that grub2 drop 30_os-prober and that its functionality be picked up by multi-boot-configtool. Actually, I would prefer to just drop the creation of these menuentry items pointing to kernels and initrds and simply chainload configuration files. Comments here would be appreciated. Timeline: I would like to see this in Fedora 21 but that is likely wishful thinking on my part. As I see them, you will be able to install both os-prober-ng and multi-boot-configtool in parallel with existing packages so they will not be on a critical path. Comments? Suggestions? BTW, I plan to attempt establishing some communication with the upstream developers/maintainers of os-prober as well as the grub2 developers to get their reaction to my proposals. Anyone wishing to be part of those discussions, please name yourselves. Gene

1 1

★ anaconda help, Nadeem left a message for you
by Nadeem 21 Jun '14

21 Jun '14

You can instantly reply using our message exchange system. Check your message http://eu1.badoo.com/nadeem.sajjad.1/in/5dHshpFajnw/?lang_id=3&g=57-0-4&m=6… Some other people in the area: If clicking the links in this message does not work, copy and paste them into the address bar of your browser. You have received this email from Badoo Trading Limited (postal address below). If you do not wish to receive further email communications from Badoo click here to opt out: https://eu1.badoo.com/impersonation.phtml?lang_id=3&email=anaconda-devel-li…. Badoo Trading Limited is a limited company registered in England and Wales under CRN 7540255 with its registered office at Media Village, 131 - 151 Great Titchfield Street, London, W1W 5BB.

1 0

Re: [PATCH 1/3] grubby: fix initrd updating when multiboot exist
by Junxiao Bi 19 Jun '14

19 Jun '14

Cc: Peter Jones <pjones(a)redhat.com> On 06/06/2014 09:59 AM, Junxiao Bi wrote: > When using the following command to add an initrd for the kernel. > grubby --update-kernel=/boot/vmlinuz-2.6.32-431.17.1.el6.x86_64.debug > --initrd /boot/initramfs-2.6.32-431.17.1.el6.x86_64.debug.img > --add-multiboot=/boot/tboot.gz > The multiboot image /boot/tboot.gz is used as the key to search the > configure entry in grub.conf, this is wrong. There may be other kernels > also configure multiboot with the same name tboot.gz, if their index are > smaller than the target one, then that will make the initrd added to the > wrong kernel. Fix it to use kernel name as the search key. > > Signed-off-by: Junxiao Bi <junxiao.bi(a)oracle.com> > --- > grubby.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/grubby.c b/grubby.c > index 88a1f08..2d7ab62 100644 > --- a/grubby.c > +++ b/grubby.c > @@ -3309,7 +3309,7 @@ int addMBInitrd(struct grubConfig * cfg, const char *newMBKernel, > > if (!image) return 0; > > - for (; (entry = findEntryByPath(cfg, newMBKernel, prefix, &index)); index++) { > + for (; (entry = findEntryByPath(cfg, image, prefix, &index)); index++) { > kernelLine = getLineByType(LT_MBMODULE, entry->lines); > if (!kernelLine) continue; >

2 4

Re: [PATCH 3/3] grubby: make kernel line same indent with hyper line
by Junxiao Bi 16 Jun '14

16 Jun '14

Cc: Peter Jones <pjones(a)redhat.com> On 06/06/2014 09:59 AM, Junxiao Bi wrote: > The configure line of kernel should have the same indent with the > one of hypervisor. The reset is no needed. > > Signed-off-by: Junxiao Bi <junxiao.bi(a)oracle.com> > --- > grubby.c | 3 --- > 1 file changed, 3 deletions(-) > > diff --git a/grubby.c b/grubby.c > index 491b07d..6e91cae 100644 > --- a/grubby.c > +++ b/grubby.c > @@ -3791,9 +3791,6 @@ int addNewKernel(struct grubConfig * config, struct singleEntry * template, > newLine = addLine(new, config->cfi, LT_HYPER, > tmplLine->indent, > newMBKernel + strlen(prefix)); > - /* set up for adding the kernel line */ > - free(tmplLine->indent); > - tmplLine->indent = strdup(config->secondaryIndent); > needs &= ~NEED_MB; > } > if (needs & NEED_KERNEL) {

1 0

Re: [PATCH 2/3] grubby: fix initrd lost issue
by Junxiao Bi 16 Jun '14

16 Jun '14

Cc: Peter Jones <pjones(a)redhat.com> On 06/06/2014 09:59 AM, Junxiao Bi wrote: > When tboot is used, there is no line with LT_KERNEL* flag in grubConfig, so > "kernelLine" will be null, this will cause updateInitrd() fail without setting > the initrd. The "kernelLine" is used here to get the initrd configure line's > type and indent, the initrd configure type should be LT_MBMODULE when tboot is > used, or it is preferredLineType(LT_INITRD, cfg->cfi). The initrd configure > indent is cfg->secondaryIndent. So remove "kernelLine". > > Signed-off-by: Junxiao Bi <junxiao.bi(a)oracle.com> > --- > grubby.c | 24 +++++++----------------- > 1 file changed, 7 insertions(+), 17 deletions(-) > > diff --git a/grubby.c b/grubby.c > index 2d7ab62..491b07d 100644 > --- a/grubby.c > +++ b/grubby.c > @@ -3340,15 +3340,12 @@ int addMBInitrd(struct grubConfig * cfg, const char *newMBKernel, > int updateInitrd(struct grubConfig * cfg, const char * image, > const char * prefix, const char * initrd) { > struct singleEntry * entry; > - struct singleLine * line, * kernelLine, *endLine = NULL; > + struct singleLine * line, *endLine = NULL; > int index = 0; > > if (!image) return 0; > > for (; (entry = findEntryByPath(cfg, image, prefix, &index)); index++) { > - kernelLine = getLineByType(LT_KERNEL|LT_KERNEL_EFI|LT_KERNEL_16, entry->lines); > - if (!kernelLine) continue; > - > line = getLineByType(LT_INITRD|LT_INITRD_EFI|LT_INITRD_16, entry->lines); > if (line) > removeLine(entry, line); > @@ -3360,21 +3357,14 @@ int updateInitrd(struct grubConfig * cfg, const char * image, > endLine = getLineByType(LT_ENTRY_END, entry->lines); > if (endLine) > removeLine(entry, endLine); > + > enum lineType_e lt; > - switch(kernelLine->type) { > - case LT_KERNEL: > - lt = LT_INITRD; > - break; > - case LT_KERNEL_EFI: > - lt = LT_INITRD_EFI; > - break; > - case LT_KERNEL_16: > - lt = LT_INITRD_16; > - break; > - default: > + if (entry->multiboot && getKeywordByType(LT_MBMODULE, cfg->cfi)) > + lt = LT_MBMODULE; > + else > lt = preferredLineType(LT_INITRD, cfg->cfi); > - } > - line = addLine(entry, cfg->cfi, lt, kernelLine->indent, initrd); > + > + line = addLine(entry, cfg->cfi, lt, cfg->secondaryIndent, initrd); > if (!line) > return 1; > if (endLine) {

1 0

disk access in %pre script
by Jerry Amundson 12 Jun '14

12 Jun '14

I'm trying to access partitions on the hard drive before the install re-formats them, but this leads to ImportError: No module named updates_disk_hook I presume because the partition is mounted by my %pre script, but the rest of kickstart continues in parallel? (In other words, not truly "pre", right? :-) Instead should I be doing something before anaconda?, e.g. http://fedoraproject.org/wiki/Anaconda/ExtendingKickstart thanks, jerry

2 2

[PATCH 1/3] grubby: fix initrd updating when multiboot exist
by Junxiao Bi 09 Jun '14

09 Jun '14

When using the following command to add an initrd for the kernel. grubby --update-kernel=/boot/vmlinuz-2.6.32-431.17.1.el6.x86_64.debug --initrd /boot/initramfs-2.6.32-431.17.1.el6.x86_64.debug.img --add-multiboot=/boot/tboot.gz The multiboot image /boot/tboot.gz is used as the key to search the configure entry in grub.conf, this is wrong. There may be other kernels also configure multiboot with the same name tboot.gz, if their index are smaller than the target one, then that will make the initrd added to the wrong kernel. Fix it to use kernel name as the search key. Signed-off-by: Junxiao Bi <junxiao.bi(a)oracle.com> --- grubby.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/grubby.c b/grubby.c index 88a1f08..2d7ab62 100644 --- a/grubby.c +++ b/grubby.c @@ -3309,7 +3309,7 @@ int addMBInitrd(struct grubConfig * cfg, const char *newMBKernel, if (!image) return 0; - for (; (entry = findEntryByPath(cfg, newMBKernel, prefix, &index)); index++) { + for (; (entry = findEntryByPath(cfg, image, prefix, &index)); index++) { kernelLine = getLineByType(LT_MBMODULE, entry->lines); if (!kernelLine) continue; -- 1.7.9.5

1 2

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Anaconda-devel June 2014