OpenCloudOS-Kernel/Documentation/admin-guide/dynamic-debug-howto.rst

377 lines
13 KiB
ReStructuredText
Raw Normal View History

Dynamic debug
+++++++++++++
Introduction
============
Dynamic debug allows you to dynamically enable/disable kernel
debug-print code to obtain additional kernel information.
If ``/proc/dynamic_debug/control`` exists, your kernel has dynamic
debug. You'll need root access (sudo su) to use this.
Dynamic debug provides:
* a Catalog of all *prdbgs* in your kernel.
``cat /proc/dynamic_debug/control`` to see them.
* a Simple query/command language to alter *prdbgs* by selecting on
any combination of 0 or 1 of:
- source filename
- function name
- line number (including ranges of line numbers)
- module name
- format string
- class name (as known/declared by each module)
Viewing Dynamic Debug Behaviour
===============================
You can view the currently configured behaviour in the *prdbg* catalog::
:#> head -n7 /proc/dynamic_debug/control
# filename:lineno [module]function flags format
init/main.c:1179 [main]initcall_blacklist =_ "blacklisting initcall %s\012
init/main.c:1218 [main]initcall_blacklisted =_ "initcall %s blacklisted\012"
init/main.c:1424 [main]run_init_process =_ " with arguments:\012"
init/main.c:1426 [main]run_init_process =_ " %s\012"
init/main.c:1427 [main]run_init_process =_ " with environment:\012"
init/main.c:1429 [main]run_init_process =_ " %s\012"
The 3rd space-delimited column shows the current flags, preceded by
a ``=`` for easy use with grep/cut. ``=p`` shows enabled callsites.
Controlling dynamic debug Behaviour
===================================
The behaviour of *prdbg* sites are controlled by writing
query/commands to the control file. Example::
# grease the interface
:#> alias ddcmd='echo $* > /proc/dynamic_debug/control'
:#> ddcmd '-p; module main func run* +p'
:#> grep =p /proc/dynamic_debug/control
init/main.c:1424 [main]run_init_process =p " with arguments:\012"
init/main.c:1426 [main]run_init_process =p " %s\012"
init/main.c:1427 [main]run_init_process =p " with environment:\012"
init/main.c:1429 [main]run_init_process =p " %s\012"
Error messages go to console/syslog::
:#> ddcmd mode foo +p
dyndbg: unknown keyword "mode"
dyndbg: query parse failed
bash: echo: write error: Invalid argument
If debugfs is also enabled and mounted, ``dynamic_debug/control`` is
also under the mount-dir, typically ``/sys/kernel/debug/``.
Command Language Reference
==========================
At the basic lexical level, a command is a sequence of words separated
by spaces or tabs. So these are all equivalent::
:#> ddcmd file svcsock.c line 1603 +p
:#> ddcmd "file svcsock.c line 1603 +p"
:#> ddcmd ' file svcsock.c line 1603 +p '
dynamic_debug: process multiple debug-queries on a line Insert ddebug_exec_queries() in place of ddebug_exec_query(). It splits the query string on [;\n], and calls ddebug_exec_query() on each. All queries are processed independent of errors, allowing a query to fail, for example when a module is not installed. Empty lines and comments are skipped. Errors are counted, and the last error seen (negative) or the number of callsites found (0 or positive) is returned. Return code checks are altered accordingly. With this, multiple queries can be given in ddebug_query, allowing more selective enabling of callsites. As a side effect, a set of commands can be batched in: cat cmd-file > $DBGMT/dynamic_debug/control We dont want a ddebug_query syntax error to kill the dynamic debug facility, so dynamic_debug_init() zeros ddebug_exec_queries()'s return code after logging the appropriate message, so that ddebug tables are preserved and $DBGMT/dynamic_debug/control file is created. This would be appropriate even without accepting multiple queries. This patch also alters ddebug_change() to return number of callsites matched (which typically is the same as number of callsites changed). ddebug_exec_query() also returns the number found, or a negative value if theres a parse error on the query. Splitting on [;\n] prevents their use in format-specs, but selecting callsites on punctuation is brittle anyway, meaningful and selective substrings are more typical. Note: splitting queries on ';' before handling trailing #comments means that a ';' also terminates a comment, and text after the ';' is treated as another query. This trailing query will almost certainly result in a parse error and thus have no effect other than the error message. The double corner case with unexpected results is: ddebug_query="func foo +p # enable foo ; +p" Signed-off-by: Jim Cromie <jim.cromie@gmail.com> Signed-off-by: Jason Baron <jbaron@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-20 06:13:21 +08:00
Command submissions are bounded by a write() system call.
Multiple commands can be written together, separated by ``;`` or ``\n``::
:#> ddcmd "func pnpacpi_get_resources +p; func pnp_assign_mem +p"
:#> ddcmd <<"EOC"
func pnpacpi_get_resources +p
func pnp_assign_mem +p
EOC
:#> cat query-batch-file > /proc/dynamic_debug/control
You can also use wildcards in each query term. The match rule supports
``*`` (matches zero or more characters) and ``?`` (matches exactly one
character). For example, you can match all usb drivers::
:#> ddcmd file "drivers/usb/*" +p # "" to suppress shell expansion
Syntactically, a command is pairs of keyword values, followed by a
flags change or setting::
command ::= match-spec* flags-spec
The match-spec's select *prdbgs* from the catalog, upon which to apply
the flags-spec, all constraints are ANDed together. An absent keyword
is the same as keyword "*".
A match specification is a keyword, which selects the attribute of
the callsite to be compared, and a value to compare against. Possible
keywords are:::
match-spec ::= 'func' string |
'file' string |
'module' string |
'format' string |
'class' string |
'line' line-range
line-range ::= lineno |
'-'lineno |
lineno'-' |
lineno'-'lineno
lineno ::= unsigned-int
.. note::
``line-range`` cannot contain space, e.g.
"1-30" is valid range but "1 - 30" is not.
The meanings of each keyword are:
func
The given string is compared against the function name
of each callsite. Example::
func svc_tcp_accept
func *recv* # in rfcomm, bluetooth, ping, tcp
file
The given string is compared against either the src-root relative
pathname, or the basename of the source file of each callsite.
Examples::
file svcsock.c
file kernel/freezer.c # ie column 1 of control file
file drivers/usb/* # all callsites under it
file inode.c:start_* # parse :tail as a func (above)
file inode.c:1-100 # parse :tail as a line-range (above)
module
The given string is compared against the module name
of each callsite. The module name is the string as
seen in ``lsmod``, i.e. without the directory or the ``.ko``
suffix and with ``-`` changed to ``_``. Examples::
module sunrpc
module nfsd
module drm* # both drm, drm_kms_helper
format
The given string is searched for in the dynamic debug format
string. Note that the string does not need to match the
entire format, only some part. Whitespace and other
special characters can be escaped using C octal character
escape ``\ooo`` notation, e.g. the space character is ``\040``.
Alternatively, the string can be enclosed in double quote
characters (``"``) or single quote characters (``'``).
Examples::
format svcrdma: // many of the NFS/RDMA server pr_debugs
format readahead // some pr_debugs in the readahead cache
format nfsd:\040SETATTR // one way to match a format with whitespace
format "nfsd: SETATTR" // a neater way to match a format with whitespace
format 'nfsd: SETATTR' // yet another way to match a format with whitespace
class
The given class_name is validated against each module, which may
have declared a list of known class_names. If the class_name is
found for a module, callsite & class matching and adjustment
proceeds. Examples::
class DRM_UT_KMS # a DRM.debug category
class JUNK # silent non-match
// class TLD_* # NOTICE: no wildcard in class names
line
The given line number or range of line numbers is compared
against the line number of each ``pr_debug()`` callsite. A single
line number matches the callsite line number exactly. A
range of line numbers matches any callsite between the first
and last line number inclusive. An empty first number means
the first line in the file, an empty last line number means the
last line number in the file. Examples::
line 1603 // exactly line 1603
line 1600-1605 // the six lines from line 1600 to line 1605
line -1605 // the 1605 lines from line 1 to line 1605
line 1600- // all lines from line 1600 to the end of the file
The flags specification comprises a change operation followed
by one or more flag characters. The change operation is one
of the characters::
- remove the given flags
+ add the given flags
= set the flags to the given flags
The flags are::
p enables the pr_debug() callsite.
_ enables no flags.
Decorator flags add to the message-prefix, in order:
t Include thread ID, or <intr>
m Include module name
f Include the function name
l Include line number
For ``print_hex_dump_debug()`` and ``print_hex_dump_bytes()``, only
the ``p`` flag has meaning, other flags are ignored.
Note the regexp ``^[-+=][flmpt_]+$`` matches a flags specification.
To clear all flags at once, use ``=_`` or ``-flmpt``.
Debug messages during Boot Process
==================================
To activate debug messages for core code and built-in modules during
the boot process, even before userspace and debugfs exists, use
``dyndbg="QUERY"`` or ``module.dyndbg="QUERY"``. QUERY follows
the syntax described above, but must not exceed 1023 characters. Your
bootloader may impose lower limits.
These ``dyndbg`` params are processed just after the ddebug tables are
processed, as part of the early_initcall. Thus you can enable debug
messages in all code run after this early_initcall via this boot
parameter.
On an x86 system for example ACPI enablement is a subsys_initcall and::
dyndbg="file ec.c +p"
will show early Embedded Controller transactions during ACPI setup if
your machine (typically a laptop) has an Embedded Controller.
PCI (or other devices) initialization also is a hot candidate for using
this boot parameter for debugging purposes.
If ``foo`` module is not built-in, ``foo.dyndbg`` will still be processed at
boot time, without effect, but will be reprocessed when module is
loaded later. Bare ``dyndbg=`` is only processed at boot.
Debug Messages at Module Initialization Time
============================================
When ``modprobe foo`` is called, modprobe scans ``/proc/cmdline`` for
``foo.params``, strips ``foo.``, and passes them to the kernel along with
params given in modprobe args or ``/etc/modprob.d/*.conf`` files,
in the following order:
1. parameters given via ``/etc/modprobe.d/*.conf``::
options foo dyndbg=+pt
options foo dyndbg # defaults to +p
2. ``foo.dyndbg`` as given in boot args, ``foo.`` is stripped and passed::
foo.dyndbg=" func bar +p; func buz +mp"
3. args to modprobe::
modprobe foo dyndbg==pmf # override previous settings
These ``dyndbg`` queries are applied in order, with last having final say.
This allows boot args to override or modify those from ``/etc/modprobe.d``
(sensible, since 1 is system wide, 2 is kernel or boot specific), and
modprobe args to override both.
In the ``foo.dyndbg="QUERY"`` form, the query must exclude ``module foo``.
``foo`` is extracted from the param-name, and applied to each query in
``QUERY``, and only 1 match-spec of each type is allowed.
The ``dyndbg`` option is a "fake" module parameter, which means:
- modules do not need to define it explicitly
- every module gets it tacitly, whether they use pr_debug or not
- it doesn't appear in ``/sys/module/$module/parameters/``
To see it, grep the control file, or inspect ``/proc/cmdline.``
For ``CONFIG_DYNAMIC_DEBUG`` kernels, any settings given at boot-time (or
enabled by ``-DDEBUG`` flag during compilation) can be disabled later via
the debugfs interface if the debug messages are no longer needed::
echo "module module_name -p" > /proc/dynamic_debug/control
Examples
========
::
// enable the message at line 1603 of file svcsock.c
:#> ddcmd 'file svcsock.c line 1603 +p'
// enable all the messages in file svcsock.c
:#> ddcmd 'file svcsock.c +p'
// enable all the messages in the NFS server module
:#> ddcmd 'module nfsd +p'
// enable all 12 messages in the function svc_process()
:#> ddcmd 'func svc_process +p'
// disable all 12 messages in the function svc_process()
:#> ddcmd 'func svc_process -p'
// enable messages for NFS calls READ, READLINK, READDIR and READDIR+.
:#> ddcmd 'format "nfsd: READ" +p'
// enable messages in files of which the paths include string "usb"
:#> ddcmd 'file *usb* +p' > /proc/dynamic_debug/control
// enable all messages
:#> ddcmd '+p' > /proc/dynamic_debug/control
// add module, function to all enabled messages
:#> ddcmd '+mf' > /proc/dynamic_debug/control
// boot-args example, with newlines and comments for readability
Kernel command line: ...
// see whats going on in dyndbg=value processing
dyndbg: refine verbosity 1-4 summary-detail adjust current v*pr_info() calls to fit an overview..detail scheme: 1- module level activity: add/remove, etc 2- command ingest, splitting, summary of effects. per >control write 3- command parsing: op, flags, search terms 4- per-site change msg can yield ~3k x 2 logs per echo "+p;-p" > command. Summarize these 4 levels in MODULE_PARM_DESC, and update verbose=3 in Doc. 2- is new, to isolate a problem where a stress-test script (which feeds ~4kb multi-command strings) would produce short writes, truncating last command and causing parsing errors, which confused test results. The script fix was to use syswrite, to deliver full proper commands. 4- gets per-callsite "changed:" pr-infos, which are very noisy during stress tests, and formerly obscured v1-3 messages, and overwhelmed the static-key workload being tested. The verbose parameter has previously seen adjustment: commit 481c0e33f1e7 ("dyndbg: refine debug verbosity; 1 is basic, 2 more chatty") The script driving these adjustments is: !/usr/bin/perl -w =for Doc 1st purpose was to benchmark the effect of wildcard queries on query performance; if wildcards are risk free cheap enough, we can deploy them in the (floating) format search. 1st finding: wildcards take 2x as long to process. 2nd purpose was to benchmark real static-key changes VS simple flag changes. Found ~100x decrease for the hard work. The script maximizes workload per >control by packing it a ~4kb string of "+p; -p;" commands; this uncovered some broken stuff. The 85th query failed, and appears to be truncated, so is gramatically incorrect. Its either an error here, or in the kernel. Its not happening atm, retest. Plot thickens: fail only happens doing +-p, not +-mf, likely load dependent. Error remains consistent. Looks like a short write, longer on writer than kernel-reader. Try syswrite on handle to control this. That fixed short write. =cut use Getopt::Std; getopts('vN:k:', \my %opts) or die <<EOH; $0 options: -v verbose -k=n kernel dyndbg verbosity -N=n number of loops.. tbrc EOH $opts{N} //= 10; # !undef, 0 tests too long. my $ctrl = '/proc/dynamic_debug/control'; vx($opts{k}) if defined $opts{k}; # works on -k0 open(my $CTL, '>', $ctrl) or die "cant open $ctrl for writing: $!\n"; sub vx { my $arg = shift; my $cmd = "echo $arg > /sys/module/dynamic_debug/parameters/verbose"; system($cmd); warn("vx problem: rc:$? err:$! qry: $cmd\n") if ($?); } sub qryOK { my $qry = shift; print "syntax test: <\n$qry>\n" if $opts{v}; my $bytes = syswrite $CTL, $qry; printf "short read: $bytes / %d\n", length $qry if $bytes < length $qry; if ($?) { warn "rc:$? err:$! qry: $qry\n"; return 0; } return 1; } sub build_queries { my ($cmd, $flags, $ct) = @_; # build experiment and reference queries my $cycle = " $cmd +$flags # on ; $cmd -$flags # off \n"; my $ref = " +$flags ; -$flags \n"; my $len = length $cycle; my $max = int(4096 / $len); # break/fit to buffer size $ct |= $max; print "qry: ct:$max x << \n$cycle >>\n"; return unless qryOK($ref); return unless qryOK($cycle); my $wild = $cycle x $ct; my $empty = $ref x $ct; printf "len: %d, %d\n", length $wild, length $empty; return { trial => $wild, ref => $empty, probe => $cycle, zero => $ref, count => $ct, max => $max }; } my $query_set = build_queries(' file "*" module "*" func "*" ', "mf"); qryOK($query_set->{zero}); qryOK($query_set->{probe}); qryOK($query_set->{ref}); qryOK($query_set->{trial}); use Benchmark; sub dobatch { my ($cmd, $flags, $reps, $ct) = @_; $reps ||= $opts{N}; my $qs = build_queries($cmd, $flags, $ct); timethese($reps, { wildcards => sub { syswrite $CTL, $qs->{trial}; }, no_search => sub { syswrite $CTL, $qs->{ref}; } } ); } sub bench_static_key_toggle { vx 0; dobatch(' file "*" module "*" func "*" ', "mf"); dobatch(' file "*" module "*" func "*" ', "p"); } sub bench_verbose_levels { for my $i (0..4) { vx $i; dobatch(' file "*" module "*" func "*" ', "mf"); } } bench_static_key_toggle(); __END__ Heres how the test-script runs: :: verbose=3 parsing info [ 48.401646] dyndbg: query 95: "file "*" module "*" func "*" -mf # off " mod:* [ 48.402040] dyndbg: split into words: "file" "*" "module" "*" "func" "*" "-mf" [ 48.402456] dyndbg: op='-' [ 48.402615] dyndbg: flags=0x6 [ 48.402779] dyndbg: *flagsp=0x0 *maskp=0xfffffff9 [ 48.403033] dyndbg: parsed: func="*" file="*" module="*" format="" lineno=0-0 [ 48.403674] dyndbg: applied: func="*" file="*" module="*" format="" lineno=0-0 :: verbose=2 >control summary. ~300k site matches/changes per 4kb command [ 48.404063] dyndbg: processed 96 queries, with 296160 matches, 0 errs :: 2 queries against each other, no-search vs all-wildcard-search qry: ct:48 x << file "*" module "*" func "*" +mf # on ; file "*" module "*" func "*" -mf # off >> len: 4080, 576 Benchmark: timing 10 iterations of no_search, wildcards... no_search: 0 wallclock secs ( 0.00 usr + 0.03 sys = 0.03 CPU) @ 333.33/s (n=10) (warning: too few iterations for a reliable count) wildcards: 0 wallclock secs ( 0.00 usr + 0.09 sys = 0.09 CPU) @ 111.11/s (n=10) (warning: too few iterations for a reliable count) :: 2 queries, both doing real work / changing stati-key states. qry: ct:49 x << file "*" module "*" func "*" +p # on ; file "*" module "*" func "*" -p # off >> len: 4067, 490 Benchmark: timing 10 iterations of no_search, wildcards... no_search: 20 wallclock secs ( 0.00 usr + 20.36 sys = 20.36 CPU) @ 0.49/s (n=10) wildcards: 21 wallclock secs ( 0.00 usr + 21.08 sys = 21.08 CPU) @ 0.47/s (n=10) bash-5.1# Thats 150k static-key-toggles / sec ~600x slower than simple flags on qemu --smp 3 run Signed-off-by: Jim Cromie <jim.cromie@gmail.com> Link: https://lore.kernel.org/r/20211019210746.185307-1-jim.cromie@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-10-20 05:07:46 +08:00
dynamic_debug.verbose=3
// enable pr_debugs in the btrfs module (can be builtin or loadable)
btrfs.dyndbg="+p"
// enable pr_debugs in all files under init/
// and the function parse_one, #cmt is stripped
dyndbg="file init/* +p #cmt ; func parse_one +p"
// enable pr_debugs in 2 functions in a module loaded later
pc87360.dyndbg="func pc87360_init_device +p; func pc87360_find +p"
Kernel Configuration
====================
Dynamic Debug is enabled via kernel config items::
CONFIG_DYNAMIC_DEBUG=y # build catalog, enables CORE
CONFIG_DYNAMIC_DEBUG_CORE=y # enable mechanics only, skip catalog
If you do not want to enable dynamic debug globally (i.e. in some embedded
system), you may set ``CONFIG_DYNAMIC_DEBUG_CORE`` as basic support of dynamic
debug and add ``ccflags := -DDYNAMIC_DEBUG_MODULE`` into the Makefile of any
modules which you'd like to dynamically debug later.
Kernel *prdbg* API
==================
The following functions are cataloged and controllable when dynamic
debug is enabled::
pr_debug()
dev_dbg()
print_hex_dump_debug()
print_hex_dump_bytes()
Otherwise, they are off by default; ``ccflags += -DDEBUG`` or
``#define DEBUG`` in a source file will enable them appropriately.
If ``CONFIG_DYNAMIC_DEBUG`` is not set, ``print_hex_dump_debug()`` is
just a shortcut for ``print_hex_dump(KERN_DEBUG)``.
For ``print_hex_dump_debug()``/``print_hex_dump_bytes()``, format string is
its ``prefix_str`` argument, if it is constant string; or ``hexdump``
in case ``prefix_str`` is built dynamically.