Resolves: #2177948, handle multiple platform dumps
This commit is contained in:
Than Ngo
parent
0c74efef2c
commit
b200305443
100
ppc64-diag-2.7.9-handle_multiple_platform_dumps.patch
Normal file
100
ppc64-diag-2.7.9-handle_multiple_platform_dumps.patch
Normal file
@ -0,0 +1,100 @@
|
||||
commit d05654e5ec6f37cf6caa491fc7d95b336f9603e2
|
||||
Author: Sathvika Vasireddy <sv@linux.ibm.com>
|
||||
Date: Mon Jul 10 13:43:21 2023 +0530
|
||||
|
||||
rtas_errd: Handle multiple platform dumps
|
||||
|
||||
Currently, whenever a new dump arrives, old dump file of that specific dump
|
||||
type is removed before writing the new dump out. Any dump file with the
|
||||
same prefix (dump type) gets deleted. This means only one set of dump files
|
||||
is saved, since only one dump file per dump type is saved.
|
||||
|
||||
Handle multiple dumps on Linux by allowing as many dumps to be offloaded
|
||||
until disk space is available. To do this, remove the function that checks
|
||||
for prefix size and removes old dump files. In the event of not enough
|
||||
disk space available, log an error to the user along with the dump tag.
|
||||
User will free up space and run extract_platdump tool using the dump tag
|
||||
provided in the error message to offload the dump. Error log can be viewed
|
||||
by the user by issuing 'journalctl -p err -t rtas_errd' command.
|
||||
|
||||
Signed-off-by: Sathvika Vasireddy <sv@linux.ibm.com>
|
||||
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
|
||||
|
||||
diff --git a/rtas_errd/dump.c b/rtas_errd/dump.c
|
||||
index cc50d91..494c322 100644
|
||||
--- a/rtas_errd/dump.c
|
||||
+++ b/rtas_errd/dump.c
|
||||
@@ -30,8 +30,10 @@
|
||||
#include <fcntl.h>
|
||||
#include <librtas.h>
|
||||
#include <librtasevent.h>
|
||||
+#include <syslog.h>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/wait.h>
|
||||
+#include <sys/statvfs.h>
|
||||
#include "utils.h"
|
||||
#include "rtas_errd.h"
|
||||
|
||||
@@ -284,7 +286,9 @@ void
|
||||
check_platform_dump(struct event *event)
|
||||
{
|
||||
struct rtas_dump_scn *dump_scn;
|
||||
+ struct statvfs vfs;
|
||||
uint64_t dump_tag;
|
||||
+ uint64_t dump_size;
|
||||
char filename[DUMP_MAX_FNAME_LEN + 20], *pos;
|
||||
char *pathname = NULL;
|
||||
FILE *f;
|
||||
@@ -306,11 +310,34 @@ check_platform_dump(struct event *event)
|
||||
return;
|
||||
}
|
||||
|
||||
- /* Retrieve the dump */
|
||||
+ /* Retrieve the dump tag */
|
||||
dump_tag = dump_scn->id;
|
||||
dump_tag |= ((uint64_t)dump_scn->v6hdr.subtype << 32);
|
||||
dbg("Dump ID: 0x%016LX", dump_tag);
|
||||
|
||||
+ if (statvfs(d_cfg.platform_dump_path, &vfs) == -1) {
|
||||
+ log_msg(event, "statvfs() failed on %s: %s",
|
||||
+ d_cfg.platform_dump_path, strerror(errno));
|
||||
+ return;
|
||||
+ }
|
||||
+
|
||||
+ /* Retrieve the size of the platform dump */
|
||||
+ dump_size = dump_scn->size_hi;
|
||||
+ dump_size <<= 32;
|
||||
+ dump_size |= dump_scn->size_lo;
|
||||
+
|
||||
+ /* Check if there is sufficient space in the file system to store the dump */
|
||||
+ if (vfs.f_bavail * vfs.f_frsize < dump_size) {
|
||||
+ syslog(LOG_ERR, "Insufficient space in %s to store platform dump for dump ID: "
|
||||
+ "0x%016lX (required: %lu bytes, available: %lu bytes)",
|
||||
+ d_cfg.platform_dump_path, dump_tag, dump_size,
|
||||
+ (vfs.f_bavail * vfs.f_frsize));
|
||||
+ syslog(LOG_ERR, "After clearing space, run 'extract_platdump "
|
||||
+ "0x%016lX'.\n", dump_tag);
|
||||
+ return;
|
||||
+ }
|
||||
+
|
||||
+ /* Retrieve the dump */
|
||||
snprintf(tmp_sys_arg, 60, "0x%016LX", (long long unsigned int)dump_tag);
|
||||
system_args[0] = EXTRACT_PLATDUMP_CMD;
|
||||
system_args[1] = tmp_sys_arg;
|
||||
diff --git a/rtas_errd/extract_platdump.c b/rtas_errd/extract_platdump.c
|
||||
index fbe65b2..831e57e 100644
|
||||
--- a/rtas_errd/extract_platdump.c
|
||||
+++ b/rtas_errd/extract_platdump.c
|
||||
@@ -290,12 +290,6 @@ extract_platform_dump(uint64_t dump_tag)
|
||||
}
|
||||
}
|
||||
|
||||
- /*
|
||||
- * Before writing the new dump out, we need to see if any older
|
||||
- * dumps need to be removed first
|
||||
- */
|
||||
- remove_old_dumpfiles(filename, prefix_size);
|
||||
-
|
||||
/* Copy the dump off to the filesystem */
|
||||
pathname[0] = '\0';
|
||||
strcpy(pathname, d_cfg.platform_dump_path);
|
||||
@ -1,6 +1,6 @@
|
||||
Name: ppc64-diag
|
||||
Version: 2.7.9
|
||||
Release: 1%{?dist}
|
||||
Release: 2%{?dist}
|
||||
Summary: PowerLinux Platform Diagnostics
|
||||
URL: https://github.com/power-ras/%{name}
|
||||
License: GPLv2
|
||||
@ -34,6 +34,10 @@ Source5: rtas_errd.8
|
||||
# fix paths and permissions
|
||||
Patch0: ppc64-diag-2.7.9-fedora.patch
|
||||
|
||||
# upstream fixes
|
||||
# rtas_errd: Handle multiple platform dumps
|
||||
Patch10: ppc64-diag-2.7.9-handle_multiple_platform_dumps.patch
|
||||
|
||||
%description
|
||||
This package contains various diagnostic tools for PowerLinux.
|
||||
These tools captures the diagnostic events from Power Systems
|
||||
@ -169,6 +173,9 @@ if [ "$1" = "0" ]; then # last uninstall
|
||||
fi
|
||||
|
||||
%changelog
|
||||
* Thu Jul 20 2023 Than Ngo <than@redhat.com> - 2.7.9-2
|
||||
- Resolves: #2177948, handle multiple platform dumps
|
||||
|
||||
* Tue Oct 18 2022 Than Ngo <than@redhat.com> - 2.7.9-1
|
||||
- Resolves: #2110128, add NVMe Call Home support
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user