There is an issue when two different files in a newer release have same hash:
This is, when swupd updates a file by applying a delta, it takes the HASH2 to
know the file where delta must be applyed. If files are different in current
release (before updating), there must be 2 deltas, one for each file but as
the two files have same hash in new release delta's name are the same:
FROM-TO-HASH2 and it is when issue arises due to just the last delta file is
kept when swupd server creates files and packs. So when swupd client tries to
apply the delta to one of the file that does not corresponds it will fail and
generates an error. At the first look it will seem like delta file is corrupted
however the issue is that delta file was created for another file.
To solve this issue we include the hash for the original file in the delta's
name so that swupd client can take the correct delta and apply it:
FROM-TO-HASH1-HASH2.
Signed-off-by: Jose R Guzman <jose.r.guzman.mosqueda@intel.com>
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
Calculate the hash for directories is desiderable to be independent
on the dirname due to the subsequent calculation on the staged/HASH
file. Here is used const "DIRECTORY" string for input name for
all folders.
Signed-off-by: Jose R Guzman <jose.r.guzman.mosqueda@intel.com>
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
Delta manifests are handled by the pack creation step, and should not be
part of the plain manifest/update content creation step. This reduces the time
swupd_create_update takes to run significantly.
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
This reverts commit 7f3b760430 and commit
74f4afa655. The code introduced parallel manifest
creation, and removed delta manifests from being created, but needs some more
rigorous testing and refactoring to ensure all test suites pass.
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
I had neglected to rebase to the latest revision of #21, so this commit
incorporates the additional changes.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
The code previously created all the bundle manifests and delta manifests
synchronously in 1 thread. While this worked, it was not optimal considering
the manifests can be created at the same time since they are unique. This
patch creates a threadpool that processes each bundle manifest in its own
thread. Delta manifest creation was removed from create_update because it
can already be handled by our delta pack creation process, and it is a huge
time-sink for create_update as it also runs single threaded.
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
The delta manifests were being staged correctly but were not being added
to the delta packs.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
There are new functions that replace system() calls: system_argv(), system_argv_fd and system_argv_pipe().
They are based on execvp() function and they can manage any argument including filenames with characters
that could be interpreted as meta-characters in a regular system() function.
So no escape for the arguments needed and is less prone to errors.
Signed-off-by: Jose R Guzman <jose.r.guzman.mosqueda@intel.com>
Since delta pack creation is very much tied to a DevOps workflow, it
makes more sense for the pack_maker.sh to live outside of swupd-server.
This script will move to the clearlinux/mixer-tools repo, since the
Clear mixer is a DevOps-oriented tool.
Also, update the code comment in versions.c to describe another action
to be taken that is DevOps-specific, namely that the value of
WEBDIR/version/formatN/latest file should be set at release time.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Just in case we encounter a bundle manifest that does not exist anymore,
or the last_change is 0, which is an invalid version to create, make
sure we add the deleted entry in the MoM and do not attempt to untar
that manifest.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
The bundle manifest hashes were calculated from the STATE_DIR Manifest.bundle
files, which caused a problem when they the permissions/owner were changed. The
hashes should be calculated by untarring the Manifest.BUNDLE.tar files
referenced in the Manifest.MoM.
Signed-off-by: Tudor Marcu <tudor.marcu@intel.com>
To better track which files are excluded from manifests because they
contain blacklisted characters, the full path is more useful to log
rather than the file basename.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Since the source files for entry points to the other two binaries are
named after the binaries, do the same for swupd_create_update.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Since removing of all of the matching deleted files between a build with
an older manifest format and current needs to happen *after*
match_manifests() finishes accounting for new and deleted entries in the
manifests, simply move the logic into a separate function to be called
after match_manifests().
Because one or more file deletions should trigger a fresh bundle manifest
creation, account for the return value of this new function in addition
to the result of manifest pruning and whether any bundle includes were
changed. Since "os-core" is treated specially, and it's expected to
contain changed files for every build, only the non-"os-core" bundle
manifests need the trigger.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
For the new full manifest and new bundle manifests, the format member
was not being initialized.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
To decrease chances of specifying an incorrect format number, since the
default is a hardcoded value "3", always require the user to pass the -F
option to swupd_create_update.
In the future, it would be nice to read the default value from a config
file, with the -F option overriding the value.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
To better support a workflow for incrementing the format on the command
line, it's useful to have the format be an integer instead of a string.
When reading a manifest (ie: old manifest) we can read that into struct
manifest. Then when building a new manifest we can compare against the
prior manifest's format. If the format incremented, we can prune files
previously marked as deleted in the old manifest so they no longer
appear in the new manifest.
This is beneficial because a file that has a version number in its name
will repeatedly be deleted and replaced by a new similar file. Over
time this leads to manifests getting bigger and bigger because they
retain the entire history of all file names which ever existed, even if
a huge number of them are and have been for a long time marked as
deleted.
Signed-off-by: Tim Pepper <timothy.c.pepper@linux.intel.com>
When deciding if a new manifest should be generated for a new version,
the included manifests were not being used to detect a manifest
change. This caused manifests that had new includes: lines in the new
version to not be generated resulting in files from the included
manifest to be missed when adding/updating that bundle.
The fix was to compare includes: lines from the current and previous
manifest versions for differences and generate new manifests when that
was the case. In order to compare manifests duplicate includes: lines
are not allowed and previously os-core could be added multiple times to
a manifest because it was always added. This change also added detection
for os-core before automatically adding it to a bundle.
Now that much of the includes plumbing present on the output/subtract
side, make sure all non os-core manifests explicitly include os-core,
and then use the new extra semantics of subtract_manifests() to subtract
all included bundles, not just os-core.
Using the $bundle-includes files for building submanifests from
directories and previous manifest files includes: sections, create the
manifest->includes list as the names of submanifests included in the
submanifest being instantiated.
This needs to then be processed into a list of pointers to the other
submanifests themselves once all other manifests have been
instantiated before use with functions that operate on the includes list
(subtract_manifest and write_manifest for now)..
When subtracting a manifest from another, all the included manifests
also will need to be subtracted from the original. To do this build up a
unique list of all manifests included (directly or indirectly) from the
manifest being subtracted from.
With that in place, this gives an opportunity to provide the
functionality of subtracting all included manifests from the main
manifest by calling subtract_manifests(A, A);
If a bundle includes one or more other bundles, print out one or more
includes: lines as part of the header of the manifest file.
There will be as many includes: lines as included bundles (including
being omitted if a bundle does not include other bundles).
To capture all information that is logged, change init_log() to make
every log file name unique, and add a call to swupd_make_fullfiles to
enable logging.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
A big chunk of time during pack creation is spent in untar'ing of
fullfiles tar files. However, we also (most likely) have this exact
same content already in the image/ directory.
This patch makes the pack creator skip the untar if its possible to link
the file directly, and falls back to the existing behavior if the link
fails
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
In subtract_manifests(), if the two files under consideration have the
same filename but should not be subtracted, a file is skipped in the m2
manifest because there is no "continue" statement at the end of the
first "if" block.
This results in potentially many files not getting subtracted when they
ought to be.
To clarify the logic, avoid using "continue", since it is not needed.
Convert to using one "if" block instead.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Even though swupd_make_pack will exit immediately in the event this
error occurs, calling free() here will be needed if the error handling
ever changes (e.g. returning from the function instead of exiting).
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Because pack generation is a time-consuming task, running this task in
parallel is preferable.
However, if we are generating two or more packs with the same "from"
version at the same time, the multiple jobs will use the same pack
staging directory, leading to corruption.
My solution for now is to use a staging dir with a name derived from
both "from" and "to" versions so that there are no collisions.
A long-term fix will be to implement some kind of locking mechanism to
prevent *identical* packs from being generated simultaneously, but for
now, using more unique staging directories will improve the situation.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Since GNU tar fails to extract files with xattrs preserved when
Integrity Measurement Architecture (IMA) is enabled some vendors
may choose to install libarchive-based tar (bsdtar) on their embedded
devices, so the swupd server needs to be able to create archives
in its format.
This patch adds one compile-time options --enable-bsdtar that is used
to enable/disable GNU tar specific options. Also it harmonizes
the command strings to be compatible with both GNU tar and bsdtar.
Particularly it
- changes --exclude pattern from '%s'/* to more explicit '%s/?*' because
bsdtar's pattern matching is greedier than in tar: it uses tcsh's
globbing where '*' can be anything including the null string and the
original pattern would include the directory itself;
- OS file names are escaped with leading ./ to avoid collisions with
file names starting with @ which has special meaning in bsdtar.
Signed-off-by: Dmitry Rozhkov <dmitry.rozhkov@linux.intel.com>
Similar to how swupd_create_update handles argc/argv, swupd_make_pack
and swupd_make_fullfiles should validate the argument count before
checking for root privileges, etc.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Currently swupd-client includes xattrs to hash sums of all files.
Therefore in order to avoid relaxing security and to prevent
`swupd verify` from reporting hash mismatches for updated files
it's better to include xattrs to hashes for all files.
Signed-off-by: Dmitry Rozhkov <dmitry.rozhkov@linux.intel.com>
Some of the existing comments cause clang-format to reformat them
strangely, so modify them to something more expected by the tool.
Also, the tool may decide to "join" several lines of concatenated string
literals if a line begins with a macro string literal. That's not
wanted, so fix up that case too.
Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>
Tudor Marcu