Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update flash update to be useable at runtime. #297

Merged
merged 15 commits into from Jul 5, 2016
Merged

Update flash update to be useable at runtime. #297

merged 15 commits into from Jul 5, 2016

Conversation

mdmillerii
Copy link
Contributor

@mdmillerii mdmillerii commented May 24, 2016 •
edited

The underlying storage for a file system can not be updated when it is in use. The initial solution was to update the flash as the system was being shutdown, with output to the console.

Updates were made to the init script that allows the flash to not be used for a given boot but the update script was not enhanced to make this user friendly, and no documentation was provided to prepare the system for this mode.

This series partially addresses #293 by enhancing the update script to

  1. provide more fine grained checks on update conflicts allowing use at runtime
  2. provide additional checks that would cause failure to flash
  3. provide user help and consistent tagging of error output
  4. provide an error return code if the checks fail
  5. provide a nice diagnostic if no images were presented for update (and not generate an error).

Two patches in this series provide fixes #378 in different ways:

  1. Correct the code to address the issue
  2. Avoid the issue by changing the line in the whitelist

In addition the series provides

  • A fix to properly assemble the flash image, filling with 0xff characters as intended.
  • A cleanup to the build recipe and naming.
  • A diagnostic message to shutdown if the update did not remove a file indicating it was successful.
  • A method to build init-options-base into the initramfs providing options to init while allowing additional sources of options like the kernel command line.
  • A self contained image initramfs-netboot.cpio.u-boot for loading over the network.

These last two are intended to ease creating and using images for alternate network based deployment and testing, using the features and option parsing already in the init script.

This change is Reviewable

This update addresses several review comments and fixes some spelling errors. The changed
lines compared to the previous version were exercised.

@mdmillerii mdmillerii mentioned this pull request Jun 15, 2016
Closed
@mdmillerii mdmillerii mentioned this pull request Jun 21, 2016
Closed
@mdmillerii
image_types_uboot: Don't truncate when assembling flash
59761f3 
The generate_flash_image step was creating a file then using dd to
fill in the pieces, but missed adding the conv=notrunc flag, so each
step was shortening the file to its output.

Add the forgotten conversion flag to each dd command.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: Fix recipe, remove unnecessary class
01d3756 
Because the recipe was named with trailing part init instead
of initfs, the package didn't match the package name requiring
PROVIDES, RPROVIDES and other bitbake workarounds.  Fix the recipe
name to match the package removing the overly verbose packaging.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: shutdown: comment on update success
089d5df 
Print an error from the shutdown script when the update fails.

Update is changing to return codes when its preconditions are not
met, so be more verbose about its success to provide a means to
debug flash updates not occurring.

We could add a sulogin or sushell but its not clear what would be
desired here.  If so the message should include "rmdir /oldroot
to get a PID 1 shell" like /takever works in init.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: Don't exec sh or sulogin on error just exit 1
e1cbebe 
When update was written it was exec'd from the shutdown script
and hence took over pid 1.  Since exiting in that environment was
a panic situation, the script instead started a rescue shell with
its output presumably on the console.

The calling convention was updated to be a simple invocation in
commit dbacf10 ("obmc-initfs: run update as a sub-script")
but the error handling was not updated.  That error handling is
now becoming a hindrance to use from additional environments so
change it.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: Do not cause an error if no files were saved
a67cc44 
If there were no files in the read-write overlay file system that
were in the persistent file list, then the save directory will
not be created and therefore not exist.  Skip attempting copying
the non-existent directory in that case to avoid the error message.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: Consistently add ERROR: and print to stderr
d16a7b0 
Make sure all error messages start with the tag ERROR and its
printed to stderr by creating an echoerr function.

Previously one case had the file descriptors backwards in the
redirection.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: add option to perform checks but skip actual update
4015b7a 
Provide a way to check if update is likely to succeed without
attempting actual updates.

This will check the image names correspond to mtd partition names,
and will be enhanced with additional checking for file size and
individual partition mounts.

This is not called --dry-run because it will still do file save
and restore if those are enabled.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: check individual mtd partitions for mounts
810dc70 
Instead of checking for any mtd device, only check mtd devices
that are to be updated.  Include child mtd devices when looking
for mounted filesystems.

If this check is suppressed flash updates are suppressed.
Suppressing the check may be useful to verify a update could
succeed at shutdown.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: add check for image size vs mtd size
1944aac 
The flashcp command will check the file size vs the partition size,
so add a check when looking at images.

Use stat -L -c "%s" to get the file size, and compare it to the mtd
size from the sysfs size attribute.

If this check is suppressed the enforcement is left to flashcp.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: add --help message
30137fa 
Now that update is intended for use at runtime and not just by
the init and shutdown scripts, add a short help message listing
the options so people don't have to understand how to read shell.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: init: add file for base options
49ce6a1 
Allow an initrd image to include options for init while still allowing
additional options to come from the kernel command line and/or u-boot
environment variables.

This will allow building a netboot image that can specify to run
from RAM while allowing the command line to initiate debug or skip
copying files from the read-write overlay file system.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
image-types_uboot: Package a self contained netboot image
f19c04b 
Create a cpio containing the base read-only image, and create a
single combined initramfs image consisting of the image-rofs
and the existing initramfs.

There has been a desire to have a self contained netboot image present
for developer testing and experimentation.   The init script already
supports having the rofs layer sourced from a file packaged into the
initramfs, and will use it if found.

While this image will not fit in the current flash layout in
the initrd space, it will load over the network.  This method
leaves the squashfs compressed in RAM, decompreses is invoked as
needed at runtime, trading boot time and memory space for runtime
decompress on demand.  This will also give similar file access
overheads and performance to reading the data from flash.

Because the rofs is already compressed with xz compression,
don't try to compress this cpio (the cpio packaging adds about 40
bytes of ascii before and after, plus some alignment padding).
Because the kernel requires uncompressed cpios to appear on a
4-byte boundary with 0 byte fill between archives, place this
cpio first in the combined image.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: handle no images to update
4e87861 
Notify the user if image wildcard expansion fails instead of printing
message about failing to find partition to flash.

The update script errors with the message that it can't figure out what
partition to flash for /run/initramfs/image-* if there are no images
pending.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: update: Sanitize whitelist directory entries
adeceff 
Repeatedly strip trailing "/" and "/." from whitelist entries
and fail if an entry includes "/../", ends with "/..", or doesn't
start with a "/".  Also use the entries quoted to avoid any glob.

It was noticed the save code was saving directories that ended
in "/" into a subdirectory of the last component name.  This was
traced the the code creating the directory just stripping the last
"/" and then copying to the directory.

Choose to sanitize the entry where possible for ease of use verses
a small performance penalty.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii
initfs: whitelist: Sanitize obmc events whitelist entry
a6e6d0c 
Remove the trailing / so that the update script doesn't have to loop.

It was determined the trailing slash confused the code that makes
the save subdirectory.  The update script was fixed but this removes
the extra work needed.

Signed-off-by: Milton Miller <miltonm@us.ibm.com>
@mdmillerii mdmillerii mentioned this pull request Jun 30, 2016
Merged
@williamspatrick williamspatrick merged commit 66859fe into openbmc:master Jul 5, 2016
spinler pushed a commit to spinler/openbmc that referenced this pull request Nov 15, 2022
@rfrandse
pldm: downstream srcrev bump 438fca9572..f39555990a (openbmc#2726)
49b1cef 
ManojKiran Eda (1):
  Fix topology links for flett and pcie switch (openbmc#297)

Reed Frandsen (1):
  Merge pull request openbmc#299 from sagisin/pldm_SAP_HANA

Matt Derksen (2):
  Support BIOS control of max freq setting (for HANA)
  Remove hana from hb_cap_hana_freq_mhz attributes

sagisin (1):
  PLDM:Wait on the Software Updater (openbmc#296)

Change-Id: I718d9703fb2bbe0a125ca51e28a7afeac9ef8728
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

events directory gets duplicated causing
2 participants
@mdmillerii @williamspatrick