Under certain conditions, e.g. in a KVM guest, a write operation
to /sys/firmware/cpi/set can result in an error message to
stderr while the script returns with a zero exit code, indicating
success.
This can cause confusion if cpictl is called by systemd due to
the existence of a writable /sys/firmware/cpi/. Let's silence
cpictl when writing to the /sys/firmware/cpi/set. This is OK, as
nothing can be inferred from failure or success from the perspective
of the invoking Linux instance.
Fixes: https://github.com/ibm-s390-linux/s390-tools/issues/120
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The maximum command line length is now dependent on the kernel image
that is loaded. Therefore move the check to check_common_ipl_data().
This function now reads the new kernel image, and check whether the
command line length is in the allowed range.
The command line size limit in zipl is now set to 64k, which is hopefully
enough.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
There are two places where we check whether the initrd and kernel image
is readably. Add one helper function that checks this. This is also a
preparation for the extended command line handling, so that we have to
add that code in only one location.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
There are quite a few places that do:
if (section == NULL) {
error_text("file '%s'", file);
} else {
error_text("file '%s' in section '%s'", file, section);
}
Add a helper function to deduplicate that.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Most of get_*_components() and finalize_*_address_data() is the same
for normal ip, tape and dump types, so we can unify that.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add a small helper function to free the common ipl data to avoid
having the same code duplicated multiple times. Also remove the if()
check before the free as it is allowed to pass NULL pointers to free().
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Each ipl type takes at least a kernel image, parameter line and initrd
parameter. This is duplicated all over the place in the current
implementation. To simplify this, add struct job_common_ipl_data which
will hold this data.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This adds support for long command lines. The boot loader stages
will no longer use a fixed limit when copying the command line, but
will use the maximum length from the kernel image that is being loaded.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The assembly code in tape0.S loads the kernel from tape to offset 0x4000.
After loading the kernel it loads the parm file and stores it to the
address provided by zipl. zipl doesn't know about the 0x4000 load offset,
so this address is inside of the kernel image and will silently corrupt
the loaded image in memory.
Fix this by copying the kernel to the final destination before loading
the parm files and initrd. We can do this because we can strip the first
0x10000 bytes like the zipl C code does for non-tape IPL.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
No need to use memcpy() and carry the length around all the time.
Switch to using strlcpy() to simplify the code.
Signed-off-by: Sven Schnelle <svens@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Compiling against fuse3 shows 'incompatible pointer type' warnings due
to the additional function arguments in the new API.
Therefore, adjust the declarations of zdsfs_getattr() and
zdsfs_readdir() to match.
GitHub-ID: https://github.com/ibm-s390-linux/s390-tools/pull/117
Signed-off-by: Graham Inggs <ginggs@debian.org>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Compiling against fuse3 shows 'incompatible pointer type' warnings due
to the additional function arguments in the new API.
Therefore, adjust the declarations of hmcdrv_fuse_getattr(),
hmcdrv_fuse_readdir() and hmcdrv_fuse_init() to match.
GitHub-ID: https://github.com/ibm-s390-linux/s390-tools/pull/117
Signed-off-by: Graham Inggs <ginggs@debian.org>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Compiling against fuse3 shows 'incompatible pointer type' warnings due
to the additional function arguments in the new API.
Therefore, adjust the declarations of cmsfs_getattr(), cmsfs_readdir(),
cmsfs_utimens(), cmsfs_rename() and cmsfs_truncate() to match.
GitHub-ID: https://github.com/ibm-s390-linux/s390-tools/pull/117
Signed-off-by: Graham Inggs <ginggs@debian.org>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When `make install` is called with parallel execution enabled (e.g.:
`-j2`), the dependencies for the necessary directories are not correct.
The `install` target depends on `install_dirs`, which creates the
necessary directories, and needs to run before any of the
`chreipl-fcp-mpath-install*` targets that have prerequisites on these
directories; but the `chreipl-fcp-mpath-install*` targets have
indirectly the same "hierarchy level" as `install_dirs`, and no direct
dependency on it.
Simplified it looks like this:
install: install_dirs chreipl-fcp-mpath-install
chreipl-fcp-mpath-install: chreipl-fcp-mpath-install-udev-rules ...
This works fine with only one job, as `install_dirs` will be satisfied
first, before recursing into any other branch of the dependency tree.
But if we have more than one job, there is nothing in the rules that
prevents Make from working on both branches - `install_dirs` and
`chreipl-fcp-mpath-install` - at the same time, and so
`chreipl-fcp-mpath-install-udev-rules` might run before `install_dirs`,
and end up with a unsatisfied dependency:
make[2]: *** No rule to make target '/usr/lib/udev/rules.d', needed by 'chreipl-fcp-mpath-install-udev-rules'. Stop.
Fix this by adding direct dependencies on `install_dirs` for all the
directory prerequisites of `chreipl-fcp-mpath-install*` targets. This
way Make will need to satisfy `install_dirs` before it can work on any
of the other targets, and so the necessary directories will be created
in the correct order.
Reported-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Steffen Maier <maier@linux.ibm.com>
Tested-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This fixes the sparse warning:
utils/crypto.c:443:68: warning: Using plain integer as NULL pointer
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This change allows mocking of print macros in unit tests.
Being able to do this in unit tests, enables us to catch output from
zgetdump functions and test it.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
_zg_err_errno() should save the current errno value before calling any
libc functions because they could change it. Failing to do so, may result
in _zg_err_errno() displaying an incorrect error message.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This change allows mocking of error/abort macros in unit tests.
Being able to do this in unit tests, enables us to test error conditions w/o
terminating the unit test runner.
The new error functions do not have "noreturn" attribute because
this would make mocking of them in unit tests impossible. We must not
compile these functions as noreturn because we need to return from them
in unit tests and returning from a noreturn function is an undefined
behavior in the C++ standard!
For more details:
- ISO/IEC 14882:2017, Chapter 10.6.8 "Noreturn attribute""
- https://en.cppreference.com/w/cpp/language/attributes/noreturn.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
It is unusual to write the file output from a tool to stdout. Therefore
add a positional argument to the zgetdump convert action where the
output file can be specified. If no positional argument is given the
output is written to stdout as before and therefore there is no change
in the default behavior.
If the file output already exists an error is returned. The reason for
this is to avoid the situation where an existing dump is accidentally
overwritten by the user.
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The keyword extern before a function declaration is not required with
modern C compilers.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The primary goal of this change is to split DFO ELF module into
multiple smaller ones for better unit testability.
The change doesn't introduce any functionality changes, just shuffling code
around.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The primary goal is to separate DFO code into multiple smaller modules
and make it unit testable.
This refactoring only moved the code around w/o changing any functionality.
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Without the fix, readers of syslog / systemd-journal can erroneously
think that dbginfo.sh was incomplete because the last logged step is
"17 of 18".
The fix also makes the missing line appear in dbginfo.log inside the
generated DBGINFO....tgz.
s390-tools v2.17.0 commit b7807d0195 ("dbginfo.sh: Add KVM commands
and rework data collection") turned create_package into a numbered
step.
Snippet of terminal output:
...
17 of 18: Postprocessing
18 of 18: Finalizing: Creating archive with collected data
Collected data was saved to:
>> /tmp/DBGINFO-....tgz <<
Please review all collected data before sending to your service organization.
Syslog before:
...
dbginfo.sh[...]: 17 of 18: Postprocessing
dbginfo.sh[...]: Data collection completed
Syslog after:
...
dbginfo.sh[...]: 17 of 18: Postprocessing
dbginfo.sh[...]: 18 of 18: Finalizing: Creating archive with collected data
dbginfo.sh[...]: Data collection completed
Fixes: b7807d0195 ("dbginfo.sh: Add KVM commands and rework data collection")
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Remove the call to `gzip` before installing the manpage during the
`make install` call. What and if compression is done should be handled by
the distribution tooling.
This also removes a dependency for the build process.
Reviewed-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
`chreipl-fcp-mpath-common.sh.in` is never executed, only used as argument
for `source` in the udev helper scripts, so the shebang is unnecessary, and
might be confusing.
Also, tools like `rpmlint` from the rpm software management will complain
about this; e.g.:
s390-tools-chreipl-fcp-mpath.s390x: W: non-executable-script /usr/lib/chreipl-fcp-mpath/chreipl-fcp-mpath-common.sh 644 /bin/bash
Reviewed-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
...to make clear why `off` is initialized to 0 and that all calls to
`dfo_chunk_add` occur for the same reason. For consistency reasons adapt
the `off` value calculation as well.
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Rename the parameter `ptr` of `notes_init` to something meaningful. Since
it points to the start of the notes segment rename it to `segment_start`.
In addition, improve the documentation for the function.
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Use `mem_chunk->size` instead of recalculate it. In addition, assign
`p_vaddr` to `p_paddr` to make clear that they must be equal and rename the
parameter `load_offset` to `elf_offset` to make clear that the offset in
the ELF vmcore file is meant. Finally rename `loads_init` to
`load_phdrs_init`.
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This cast makes the macros easier to use without getting compiler warnings.
e.g.
dfo_elf.c: In function 'dfo_elf_init':
dfo_elf.c:343:20: warning: assignment to 'Elf64_Phdr *' from incompatible pointer type 'char *' [-Wincompatible-pointer-types]
343 | phdrs_load = PTR_ADD(phdr_notes, sizeof(Elf64_Phdr) * phdrs_notes_count);
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
With gcc < 8.1 the following compile error occurs:
lshwc.c:52:37: error: initializer element is not constant
static unsigned int read_interval = ioctlsleep, cfvn, csvn, authorization;
^~~~~~~~~~
Even though this gcc bug was fixed in gcc 8.1+, change the constant
definition to a #define to make the tool compile with older compilers as
well.
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69960
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66618
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add a README.md for chreipl-fcp-mpath and infrastructure to build and
install a man page generated from the README.md. Building and installing
of the man page is only done when supplying the make option `ENABLE_DOC=1`
(per default disabled). The `pandoc` utility is used to build the man page.
The README.md is reused since there is no other tool or command as part of
chreipl-fcp-mpath that needs to be documented in a separate man page.
Also add information about chreipl-fcp-mpath to the main README.md and
CHANGELOG.md.
Reviewed-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
