From ac9c9c8c7a4e289dc46b3b128b9288caca37b683 Mon Sep 17 00:00:00 2001
From: Patrick Daly <pdaly@codeaurora.org>
Date: Wed, 8 Oct 2014 17:19:56 -0700
Subject: [PATCH] lrdp-v2: Add error recovery capablity for kernel logbuf
Requires CONFIG_LOG_BUF_MAGIC; otherwise no error checking will be
performed.
A log record is valid if:
1) The magic value matches
2) the length recorded in the header matches the sum of the lengths of the
subelements.
If a log record fails this check, perform a linear search
to find the next matching log record.
Change-Id: I584350c8bc1bf93c2d956fa1e05db8207cc2d40d
---
linux-ramdump-parser-v2/dmesglib.py | 55 +++++++++++++++++++++++++++++
1 file changed, 55 insertions(+)
diff --git a/linux-ramdump-parser-v2/dmesglib.py b/linux-ramdump-parser-v2/dmesglib.py
index a9d00e5..c22a2db 100644
--- a/linux-ramdump-parser-v2/dmesglib.py
+++ b/linux-ramdump-parser-v2/dmesglib.py
@@ -14,6 +14,8 @@ import string
from parser_util import cleanupString
+LOG_MAGIC = 0x5d7aefca
+
class DmesgLib(object):
def __init__(self, ramdump, outfile):
@@ -46,6 +48,58 @@ class DmesgLib(object):
else:
return idx + msg_len
+ def verify_log_helper(self, msg, verbose):
+ # return early if CONFIG_LOG_BUF_MAGIC is not defined
+ log_align_addr = self.ramdump.addr_lookup('__log_align')
+ if (log_align_addr is None):
+ return True
+
+ len_offset = self.ramdump.field_offset(self.struct_name, 'len')
+ text_offset = self.ramdump.field_offset(self.struct_name, 'text_len')
+ dict_offset = self.ramdump.field_offset(self.struct_name, 'dict_len')
+ magic_offset = self.ramdump.field_offset(self.struct_name, 'magic')
+ msg_len = self.ramdump.read_u16(msg + len_offset)
+ text_len = self.ramdump.read_u16(msg + text_offset)
+ dict_len = self.ramdump.read_u16(msg + dict_offset)
+ magic = self.ramdump.read_u32(msg + magic_offset)
+ log_size = self.ramdump.sizeof(self.struct_name)
+ log_align = self.ramdump.read_u32(log_align_addr)
+ is_logwrap_marker = not bool(text_len | msg_len | dict_len)
+
+ err = []
+ if (magic != LOG_MAGIC):
+ err.append('Bad Magic')
+
+ computed_msg_len = (text_len + dict_len + log_size + log_align - 1) & ~(log_align - 1)
+ if (not is_logwrap_marker and (msg_len != computed_msg_len)):
+ err.append('Bad length')
+
+ err = ' '.join(err)
+ if (err):
+ if (verbose):
+ f = '--------- Corrupted Dmesg {} for record @ {:x} ---------\n'.format(err, msg)
+ self.outfile.write(f)
+ f = self.ramdump.hexdump(msg - 0x40, 0xC0)
+ self.outfile.write(f)
+ return False
+ return True
+
+ def verify_log(self, msg, logbuf_addr, last_idx):
+ logbuf_size = self.ramdump.sizeof('__log_buf')
+ log_size = self.ramdump.sizeof(self.struct_name)
+
+ verbose = True
+ while msg != logbuf_addr + last_idx:
+ if (self.verify_log_helper(msg, verbose)):
+ return msg
+ verbose = False
+ msg = msg + 0x4
+ if (msg > logbuf_addr + logbuf_size - log_size):
+ msg = logbuf_addr
+ self.wrap_cnt += 1
+
+ return logbuf_addr + last_idx
+
def extract_dmesg_flat(self):
addr = self.ramdump.read_word(self.ramdump.addr_lookup('log_buf'))
size = self.ramdump.read_word(self.ramdump.addr_lookup('log_buf_len'))
@@ -75,6 +129,7 @@ class DmesgLib(object):
timestamp / 1000000000, (timestamp % 1000000000) / 1000, partial)
self.outfile.write(f)
curr_idx = self.log_next(curr_idx, logbuf_addr)
+ curr_idx = self.verify_log(curr_idx, logbuf_addr, last_idx)
def extract_dmesg(self):
if re.search('3.7.\d', self.ramdump.version) is not None:
--
GitLab