The library is used by multiple PV related tools, not only for managing
the guest secret store.
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Make EcPubKeyCoord available, after renaming it from EcdhPubkeyCoord.
This enables API users to calculate the hash of a public host-key and
compare it to a hash from attestation.
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Create one implementation for the verbose option to be used by all
tools. While at it, add a quiet option to decrease the verbosity.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Newer s390 kernels provide the new sysfs /sys/firmware/dump/dump_area_size
attribute which specifies the exact size of machine's dump area size.
Use this value if it is present but fall back to the previous and
slightly improved way of computation of dump area size to support
also older s390 kernels w/o the new sysfs attribute. This has
the advantage that util_arch_hsa_maxsize() has no longer to be adapted
with every introduction of yet another machine type in the future.
Suggested-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
The linear search in the nested loop over SCSI devices and over the indexed
array is still time consuming. So replace the indexed array with an
associative array. Build the array once outside of the nested loop by using
pathname expansion and extracting key and value from each item. Within
the SCSI device loop, an array entry is removed by means of the key without
linear search.
This commit is separate from the previous one so it is possible to revert
this one in case associative arrays would not be available and still get
the speed improvement from the indexed array in the previous commit.
Before:
$ time ~/git/s390-tools/zconf/lszfcp -De | wc -l
1036
real 0m17.605s
user 0m17.159s
sys 0m0.308s
After:
$ time ~/git/s390-tools/zconf/lszfcp -De | wc -l
1036
real 0m0.207s
user 0m0.175s
sys 0m0.032s
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Apparently the repeated string concatenation and word splitting was
much slower.
While at it, also remember in $ZFCP_UNIT_PATH whether the loop over
ZFCP_UNIT_ARRAY already found a zfcp_unit and re-use the path string to
replace the previous file glob construct generating the zfcp_unit path by
means of pathname expansion. This only works for the extended output case.
Before:
$ time lszfcp -De | wc -l
1036
real 2m15.387s
user 2m9.323s
sys 0m5.130s
After:
$ time ~/git/s390-tools/zconf/lszfcp -De | wc -l
1036
real 0m17.605s
user 0m17.159s
sys 0m0.308s
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Just go ahead and read the sysfs attribute directly as the syscalls are
needed anyway.
If necessary stderr could be redirected to /dev/null in the future,
but there was no case so far where the attribute did not exist.
This does not make a runtime difference when running on sysfs. Change it
anyway to avoid somebody searching for speed improvement options stumbling
over it again.
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
It's not on the hot path as the glob match is done only once.
But there is no point in matching anything but SCSI devices
such as "fc_transport" and "subsystem"
just to have an additional child process filter it once more.
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
A simple built-in parameter expansion removing the matching prefix
pattern (everything up to and including the last slash) is sufficient to
strip the path from the basename.
Speeds up "lszfcp -D" significantly.
Before:
$ time ~/git/s390-tools/zconf/lszfcp -D | wc -l
1028
real 0m5.322s
user 0m0.300s
sys 0m4.907s
After:
$ time ~/git/s390-tools/zconf/lszfcp -D | wc -l
1028
real 0m0.602s
user 0m0.100s
sys 0m0.492s
Above measurements are from the fast in-memory sysfs. The improvements are
likely amplified when running "lszfcp --sysfs ..." on collected debug data
in a regular file system.
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
An online CCW device, which is available and not defunct and not failed,
can have an undesired fc_host port_state, which is not Online and does not
have the special case Linkdown. Indicate such unexpected state.
Complements
v2.9.0 commit 4036e80b26 ("lszfcp: add new output marker for non-good FCP
devices (hosts)")
v2.9.0 commit be7b854969 ("lszfcp: also mark FCP devices (hosts) that are
not available")
v2.18.0 commit 45e3f016f4 ("lszfcp: add linkdown case to host marker of
extended output")
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Switch from atoi() to existing parse_int() helper function for parsing
the integer --cmg argument to better handle invalid values such as
numbers followed by an unsupported suffix (e.g. 1x).
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Vineeth Vijayan <vneethv@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
At the moment the memory range of processed vmdump always corresponds to
the highest defined byte (q v storage) obtained from Address Space
Information Block ignoring memory ranges specified in CP vmdump command.
This might lead to the invalid dump size calculation upon conversion to
s390 format.
Use Requested Range Table from ASIZBK to identify actual memory ranges
being dumped. Consider the end of the highest memory range as a vmdump
upper limit (dump memory size). In case no range table entries present
fall back to the original method using 'storage_size_def_store' value.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
This patch adds support for zipl targets over partitions of md-devices.
With this patch it is possible to specify a zipl target over any
partition of an md-device, so that all the physical disks participating
in the mirrored setup will be prepared for IPL.
Unlike a whole md-device which has major number 9, its partitions
have major numbers 259 and the driver name is identified as 'blkext'
in '/proc/devices'. Handle this case in set_driver_name(): call ioctl()
to make sure that device is an md-partition.
Drop re-definition of some macros.
Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Add "zhypinfo" as an overview on Hypervisor structure in runtime.out
Get an overview of all block devices in front of the detailed outputs
using "lsblk" into runtime.out
Starting OpenSSL3.0 "openssl engine" is deprecated and needs replacement by:
"openssl list --providers" (keep "engine" for older versions)
Suggested-by: Christian Rund <Christian.Rund@de.ibm.com>
Reviewed-by: Mario Held <mario.held@de.ibm.com>
Signed-off-by: Joern Siglen <siglen@de.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Define and set the minimal supported Rust version (MSRV) to 1.75.0. This
is the (current) minimal supported version over all supported distros
that use rust in s390-tools.
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
String is more or less a Vec<u8> with some extra invariants (i.e. only
UFF-8 chars). Zeroroize is implemented by calling the Vec<u8>
implementation. The zero byte is a valid UTF-8 symbol. The String
invariant is uphold by the clearing code.
Also, implement a into_inner function for clone-able inner types.
This allows converting confidential types into no-confidential types.
As Drop is implemented this requires a clone (see E0509).
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Generalize the `Confidential` impl over Vec<T> and [COUNT; T] instead of
specializing T to u8.
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Migrate Confidential to pv_core crate. This has no impact on the public
API of pv as the migrated symbols are reexported.
This enables pv_core to manage confidential data as well.
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
The size of an EC-secp521r1 signature can be up to 139 bytes (if ASN1
encoded). There is no lower bound, although sizes lower than 137 are
very unlikely but possible.
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
The new cpacfinfo command provides information about CPACF, such as which CPACF functions are installed, and make use of the new MSA 13 Query Authentication Information function.
Signed-off-by: Finn Callies <fcallies@linux.ibm.com>
Reviewed-by: Harald Freudenberger <freude@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Add extra trace logs for vmdump processing.
Log number of Index Pages, Bit-Key Pages and actual stored pages in the
vmdump.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Add a new format option to HexSlice to print every byte with a separated
space in between. Additionally the old format option 'alternate'
together with the new one results in another new format which prints
every byte separated by a space AND every byte has a leading '0x'.
Signed-off-by: Finn Callies <fcallies@linux.ibm.com>
Co-developed-by: Steffen Eiden <seiden@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
It's uncommon and prone to error to silently stop decoding/parsing a
hex-string if there is an invalid character. Therefore, add a new
function `decode_hex` which fixes this behavior and use it in the code.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Make it possible to convert internal key format to OpenSSL public key
format and vice versa. E.g. this can be useful when reading UV data
structures.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This makes it possible to reuse the constant. In addition, change the
type from i32 to usize since the value describes a size.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The size of an AES key depends on the mode of operation, so add this
information at least in the comments, as the API cannot be changed as it
is already published.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Use Self instead of the struct name whenever possible.
Automagically replace struct name with Self:
`cargo clippy --fix -- -W clippy::use_self`
This streamlines the code.
Reviewed-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When a queue is removed from a SE guest which was in a state other
than "usable" (for example "unbound") the state displayed by lszcrypt
switches to "usable" until the queue device is finally removed by the
AP bus scan running every 30s.
This intermediate state is caused by reading 0x00000000 on the
underlying /sys/devices/cardxx/xx.yyyy/ap_functions. lszcrypt only
extracts the BS bits from this value and maps these both bits to
string output
0: "usable"
1: "bound"
2: "unbound"
3: "illicit"
totally ignoring the fact that there is no AP function at all.
Now the code checks for a valid ap_functions value first, before
actually extracting and displaying the SE state. In case the
ap_functions reads as 0x00000000 lszcrypt now displays the string
"invalid".
Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
According to vmdump format, each bit of the Index page represents a Bit-Key
page and indicates if that Bit-Key page exists.
Due to the bug in vmdump64big_init() only 4096 bits of the Index Page were
processed instead of 32768. Thus for vmdumps of guests with defined storage
above 64G we could have invalid memory_start_record calculated. This ruins
the VMDUMP mem chunk read callback function.
One of the problem symptoms reported was missing UTS data in
'zgetdump -i' output caused by inability to locate OS_INFO in the dump.
Fixes: 067dd5c7ef ("zdump: Add vmdump dfi for vmdump format to elf format")
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Newer rust std libary code checks if a fd is already closes and panics.
Test code created a file with fd=17 and relied that test code never
touched the file. This cannot be done anymore. Just use the working dir
as backing file.
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
`sizeof` was added into the prelude in rustc 1.80.
This triggers a lint-warning for `unused_qualifications` if rustc 1.80+
is used. Fix this warning by using a use statement to stay compatible
for <1.80.
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
The RCs 30, 31, 32 may be caused by a malicious host or host bug.
Make this clear and advise to stop the SE-guest.
Reviewed-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
When checking CDL formatted DASD for installed dump-tool (zgetdump -d),
we might stumble upon an 'empty' boot record pointer in the volume label
filled with bytes of 0x40 (EBCDIC spaces). This leads to the following
seek error reported:
# zgetdump -d /dev/dasdb
zgetdump: Could not seek "/dev/dasdb" (Invalid argument)
Check for empty boot record and return proper error code by
ngdump_get_eckd_part_num() in order to end up with a regular
"No dump tool found" message.
Fixes: f3bcd94524 ("zdump: Add zgetdump -d support for ECKD ldipl-dump")
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
This fixes a segfault when the option '--dry-run' is specified for
device-based dumps (for which the bootmap file is created on a RAW
device, not formatted with a file system).
The actual status is that the option '--dry-run' for device-based
dumps is not implemented. The implementation is evaluated as not
trivial: In contrast with filesystem-based dumps, it is not possible
to simply make do with rename/unlink of temporary created bootmap.
The fixup returns error on any not file system based dumps being
created with the option '--dry-run'.
Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
Reviewed-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Drop a pair of duplicated assignmets
Replace the parameter 'struct job_data' with 'struct job_dump_data'
Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Move all checks from check_dump_device_late() to disk_is_approproate()
and drop check_dump_device_late() entirely
Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
In the function check_dump_device_late() drop the case of
(info->device != target_info->device) which never happens. Indeed,
the @info created in check_dump_device_late() coincides with the
@target_info previously created by
prepare_build_program_table_{file, device}.
Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
Reviewed-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Since the following kernel commit:
"32db401965f1 s390/mm: Pin identity mapping base to zero"
we can have zero identity_base value in os_info. This makes zgetdump
check for non-VR kernel in vm_info_get() incorrect what leads to missing
data (e.g. UTS kernel version) in 'zgetdump -i' output.
Check for zero kaslr_offset value instead (like crash and makedumpfile
do) in order to identify non-VR kernel.
Signed-off-by: Mikhail Zaslonko <zaslonko@linux.ibm.com>
Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
