Daniel Wagner [Fri, 17 Dec 2021 15:45:10 +0000 (16:45 +0100)]
nvme-print: Add Controller Metadata and Namespace Metadata
Port and extent the libnvme part from the commit b8a403b ("Add NVMe MI
Features: Controller Metadata (0x7E) and Namespace Metadata (0x7F).")
from the nvme-cli monolithic branch.
Daniel Wagner [Wed, 15 Dec 2021 13:22:02 +0000 (14:22 +0100)]
build: Use compiler.find_library for libhugetlbfs
libhugetlbfs is not detected via dependency() because the library
neither provides a pkg-config() nor a cmake helper. Use
compiler.find_library to figure out if the library is available.
We do not use the has_header feature of compiler.find_library()
function as this would update our minimum version for meson to
0.50.0. Instead open code it by using compiler.has_header().
Daniel Wagner [Wed, 15 Dec 2021 12:18:06 +0000 (13:18 +0100)]
build: Add wrapdb support for OpenSSL
The meson wrapdb library has support for OpenSSL. Introduce a combo
option for full control if the user wants OpenSSL added to nvme-cli or
not. Only if the option is set to 'true' the fallback will be
considered when there is no system version of OpenSSL available.
Brandon Paupore [Tue, 14 Dec 2021 17:49:15 +0000 (11:49 -0600)]
Fix JSON report zones output
Previously multiple JSON objects would be in the output of
zns-report-zones if there were more than 1024 zones requested because of
the chunked requests sent by nvme-cli.
Fix this to output a single JSON object with one zone_list. This is done
by passing around and appending to the zone_list, then outputting all of
it once after receiving all the reports.
Wen Xiong [Wed, 1 Dec 2021 22:25:58 +0000 (17:25 -0500)]
nvme-cli: Decode "Supported Events Bitmap" in PEL header
"Supported Events Bitmap" in PEL header shows what events
are supported in current nvme devices.
Persistent Event Log for device: nvme0n1
Action for Persistent Event Log: 0
..
..
..
Supported Events Bitmap:
Support SMART/Health Log Snapshot Event(0x1)
Support Firmware Commit Event(0x2)
Support Timestamp Change Event(0x3)
Support Power-on or Reset Event(0x4)
Support NVM Subsystem Hardware Error Event(0x5)
Support Change Namespace Event(0x6)
Support Format NVM Start Event(0x7)
Support Format NVM Completion Event(0x8)
Support Sanitize Start Event(0x9)
Support Sanitize Completion Event(0xa)
Support Set Feature Event(0xb)
Support Set Telemetry CRT Event(0xc)
Support Thermal Excursion Event(0xd)
Steven Seungcheol Lee [Fri, 10 Dec 2021 11:39:40 +0000 (20:39 +0900)]
zns: Add printing Zone Capacity Changed from zone reset
if the Variable Zone Capacity bit is set to ‘1’ in the Zone Operation
Characteristics field in the Zoned Namespace Command Set specific
Identify Namespace data structure, then the zone capacity may change
upon successful completion of a Zone Management Send command
specifying the Zone Send Action of Reset Zone
Signed-off-by: Steven Seungcheol Lee <sc108.lee@samsung.com>
Steven Seungcheol Lee [Mon, 22 Nov 2021 10:27:08 +0000 (19:27 +0900)]
Add version check for OpenSSL(1.1.0) required
HMAC_CTX_new(), HMAC_CTX_free() are new in OpenSSL 1.1.0.
nvme.c: In function ‘gen_dhchap_key’:
nvme.c:6792:24: error: implicit declaration of function ‘HMAC_CTX_new’; did you mean ‘HMAC_CTX_init’? [-Werror=implicit-function-declaration]
6792 | HMAC_CTX *hmac_ctx = HMAC_CTX_new();
| ^~~~~~~~~~~~
| HMAC_CTX_init
nvme.c:6792:24: error: initialization of ‘HMAC_CTX *’ {aka ‘struct hmac_ctx_st *’} from ‘int’ makes pointer from integer without a cast [-Werror=int-conversion]
nvme.c:6805:3: error: implicit declaration of function ‘HMAC_CTX_free’; did you mean ‘HMAC_CTX_copy’? [-Werror=implicit-function-declaration]
6805 | HMAC_CTX_free(hmac_ctx);
| ^~~~~~~~~~~~~
| HMAC_CTX_copy
cc1: all warnings being treated as errors
make: *** [Makefile:155: nvme.o] Error 1
Signed-off-by: Steven Seungcheol Lee <sc108.lee@samsung.com>
Steven Seungcheol Lee [Mon, 22 Nov 2021 08:34:58 +0000 (17:34 +0900)]
nvme: fix fatal error for sys/random.h
fatal error: sys/random.h: No such file or directory
sys/random.h supported from glibc version 2.25
so add header file existence check
if there is no sys/random.h, try to read /dev/urandom or /dev/random
Signed-off-by: Steven Seungcheol Lee <sc108.lee@samsung.com>
[dwagner: cache errno value, close will overwrite it] Signed-off-by: Daniel Wagner <dwagner@suse.de>
Steven Seungcheol Lee [Thu, 9 Dec 2021 11:34:24 +0000 (20:34 +0900)]
util/json: fix build error of json_object_add_value_uint64 #1260
util/json used when system does not have json-c library
json_object_add_value_uint handled with unsigned long long (64bits)
That's why json_object_add_value_uint64 is redefined with it
Signed-off-by: Steven Seungcheol Lee <sc108.lee@samsung.com>
Daniel Wagner [Thu, 9 Dec 2021 12:48:30 +0000 (13:48 +0100)]
plugins: Force types for isalnum and isblank
The two users of isalnum and isblank hand in unsigned char and not
int as expected by the API. This triggers thye type checker in the
debug build, e.g.
../subprojects/libnvme/ccan/ccan/build_assert/build_assert.h:38:22: error: size of unnamed array is negative
38 | (sizeof(char [1 - 2*!(cond)]) - 1)
| ^
../ccan/ccan/str/str.h:189:16: note: in expansion of macro ‘BUILD_ASSERT_OR_ZERO’
189 | ((i) + BUILD_ASSERT_OR_ZERO(!__builtin_types_compatible_p(typeof(i), \
| ^~~~~~~~~~~~~~~~~~~~
../ccan/ccan/str/str.h:196:32: note: in expansion of macro ‘str_check_arg_’
196 | #define isalnum(i) str_isalnum(str_check_arg_(i))
| ^~~~~~~~~~~~~~
../plugins/memblaze/memblaze-nvme.c:410:13: note: in expansion of macro ‘isalnum’
410 | if (isalnum(*c) == 0) {
| ^~~~~~~
Brandon Paupore [Thu, 2 Dec 2021 21:40:39 +0000 (15:40 -0600)]
Use unique names for VULs, also return 0 on success
Previously nvme_log_id_to_string returned "WDC Vendor Unique Log ID" if
it was any vendor unique log ID. This causes problems when requesting
JSON output, as only the final vendor unique log ID would be shown.
Instead, append the log id to the string making each unique.
Also, wdc_log_page_directory previously returned whatever the
output_format was. Instead return 0 if the format was valid.
nvme-check-dhchap-key, nvme-gen-dhchap-key the line of === under title
need to be written same length as title
-> fix [asciidoc: FAILED: manpage document title is mandatory]
nvme-connect-all, nvme-connect, nvme-discover new options are added
Signed-off-by: Steven Seungcheol Lee <sc108.lee@samsung.com>
Hannes Reinecke [Tue, 23 Nov 2021 07:19:20 +0000 (08:19 +0100)]
fabrics: do not print 'unrecognized' address family for loop
fabrics implementations like 'loop' do not have a transport address,
and consequently the address family value doesn't make any sense here.
So do not try to translate the address family to a printable string
if the address itself is empty.
Daniel Wagner [Thu, 18 Nov 2021 09:40:59 +0000 (10:40 +0100)]
build: Don't use external script to set version
The meson documentation points out the the version string should be a
fixed string. Our trick with trying to figure out the version is not
working in all cases, e.g. 'meson dist' will create a wrong file name.
The idea is to update the version string when a new development cycle
starts and mark the release with a tag. This is what other projects
do as well. Let's follow their common practice.
Daniel Wagner [Fri, 12 Nov 2021 15:29:20 +0000 (16:29 +0100)]
scripts: Replace gen-hostnqn workaround with 'nvme gen-hostnqn'
The script was introduced to workaround the short commings 'nvme
gen-hostnqn'. In the meantime these have been address. Instead
removing the script completly keep it around for backwards
compatibiliy and just use 'nvme gen-hostnqn'.
Install additional files such as dracut, systemd and udev scipts.
Also add support for building and installing documentation.
Introduce a build option '-Ddocs=[all,man,html]' which builds (if
asciidoctor is availalbe) and installs the docs.
meson doesn't support pre compiled sources updated in the source
tree. This is a design decision. That means the man pages
(Documentation/*.1) and the html pages (Documentation/*.html) can't be
updated so easily as with Makefile.
The workaround provide a shell script which updates the in tree
pre compiled documentation
Hannes Reinecke [Tue, 16 Nov 2021 13:57:05 +0000 (14:57 +0100)]
fabrics: call nvme_ctrl_set_discovery_ctrl()
When creating a discovery connection the user might be using the
unique discovery NQN, so we should be calling nvme_ctrl_set_discovery_ctrl()
to mark the connection as a discovery connection.
Hannes Reinecke [Mon, 15 Nov 2021 11:29:23 +0000 (12:29 +0100)]
nvme: document 'config-file' commandline option
The 'config-file' commandline option is not documented in the
man pages. And we should reference the config file as a 'ghost'
file in the rpm spec to ensure it gets removed together with the
nvme-cli package.
Hannes Reinecke [Wed, 17 Nov 2021 10:00:04 +0000 (11:00 +0100)]
Compability for older json-c releases
commit 0b63a8b7cfb7 ("Check json-c libversion, recover commit d127537")
broke building on older releases when using 'make'.
As the original issue was about json-c not being able to parse full 64
numbers this patch separates out 32 and 64 json objects and adds a warning
if an overflow occurs on older json-c releases.
projects / users / sagi / nvme-cli.git / log