.../admin-guide/dynamic-debug-howto.rst | 246 +++++----- MAINTAINERS | 2 + drivers/gpu/drm/Kconfig | 12 + drivers/gpu/drm/Makefile | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 14 + drivers/gpu/drm/display/drm_dp_helper.c | 13 + drivers/gpu/drm/drm_crtc_helper.c | 13 + drivers/gpu/drm/drm_print.c | 48 +- drivers/gpu/drm/i915/i915_params.c | 12 + .../gpu/drm/nouveau/include/nvkm/core/debug.h | 16 + .../drm/nouveau/include/nvkm/core/subdev.h | 17 +- drivers/gpu/drm/nouveau/nouveau_drm.c | 20 + drivers/gpu/drm/nouveau/nouveau_drv.h | 16 +- drivers/gpu/drm/nouveau/nvkm/core/subdev.c | 23 + include/asm-generic/vmlinux.lds.h | 3 + include/drm/drm_print.h | 85 +++- include/linux/dynamic_debug.h | 176 +++++-- kernel/module/internal.h | 4 +- kernel/module/main.c | 20 +- lib/Kconfig.debug | 10 + lib/Makefile | 1 + lib/dynamic_debug.c | 450 +++++++++++++++--- lib/test_dynamic_debug.c | 165 +++++++ 23 files changed, 1099 insertions(+), 269 deletions(-) create mode 100644 lib/test_dynamic_debug.c
Hi Jason, Greg, DRM-folk,
heres V5. Diffs vs V4:
1st 9 have Jason's Ack
10 is new: simple var cleanup
11 new struct, to contain future addins. touches kernel/module/
dyndbg-param callbacks moved to last (of dyndbg patches)
they're where the uncertainty is,
and the bulk of the API addition (class keyword is small)
patches between the move are the guts of the "class foo" feature.
exposure of feature is only via >control, until sysfs parts added
DRM chunk:
included here to see how patchwork's merge & test proceeds.
categories -> classes re-work is unchanged - feels pretty solid.
nouveau & LEVEL made some progress, still WIP.
I left out tracefs bits for now.
UNCERTAINTIES / RFC:
1st is my "sharing" technique, across DRM's multi modules. Its
literally duplicated decls, by each participating module, each adding
static data in its own special section. I didn't try at all to shrink
or unify this any further. There might be a perfect place to add a
single decl, but that quest will require a pony.
Sharing by class-names lacks the compiler support that varnames get;
"incomplete" enables will probably happen if one of the declarations
miss-spells one of the classnames.
That said, see the how-to below: stringifying va-args in the macro
would allow direct use of the enum-consts, and cc will insist that
those consts are defined, eliminating part of the un-support.
I could split the _BITS,_NUM handling out from _NAMES (ie split
drm.debug style patch); theres no user-demand for the _NAMES feature
ATM, and theres probably some legit API-bikeshedding to do.
PATCHSET DESCRIPTION
This patchset adds 'typed' "class FOO" support to dynamic-debug, where
'typed' means either DISJOINT_BITS, like drm_dbg(DRM_UT_CORE,..), or
LEVEL_NUM (like nouveau debug-levels). Use it in DRM modules: core,
helpers, and in drivers i915, amdgpu, nouveau.
If a module is using class'd prdbgs (pr_debug_class, dev_dbg_class, or
adapted drm_dbg_<category>) or similar in its code, it can "opt in" to
allow dyndbg to manipulate those class'd prdebugs, by declaring in a
c-file:
DECLARE_DYNDBG_CLASSMAP(drm_debug_classes,
DD_CLASS_TYPE_DISJOINT_BITS, 0,
"DRM_UT_CORE",
"DRM_UT_DRIVER",
"DRM_UT_KMS",
"DRM_UT_PRIME",
"DRM_UT_ATOMIC",
"DRM_UT_VBL",
"DRM_UT_STATE",
"DRM_UT_LEASE",
"DRM_UT_DP",
"DRM_UT_DRMRES");
// how-to stringify __va_args inside the macro ?
By doing this, a module tells dyndbg that it:
- is using class-ids [0..N] in prdbg callsites
( 0..N == DRM_UT_CORE..DRM_UT_DRMRES )
- wants to refer to them by class-names [0..N]
- is mapping those names to those class-ids
- expects users to enable them via >control or >parameter/knob
Then, a user can enable the prdbgs by their class:
:#> echo class DRM_UT_KMS +p > /proc/dynamic_debug/control
And with another 3-line bitmap param decl/init, wrapping the
drm_debug_classes var in a module-param-cb:
:#> echo 0x1 > /sys/module/drm/parameters/debug
and optionally using classnames:
:#> echo +DRM_UT_CORE,-DRM_UT_KMS \
> /sys/module/drm/parameters/debug_cats
DYNAMIC_DEBUG gets:
struct _ddebug gets a new .class_id:5 field, big enough to represent
drm_debug_category (after squeezing). It defaults to 31 for all
existing prdbgs. class_id also supports verbose uses.
classmaps (as declared by macro above) are in their own linker
section, are loaded by kernel/module, and handled by add_module,
which attaches classmaps to their module's ddebug table.
ddebug_change() handles a class FOO query by validating that FOO is
known by each module in the loop. The query is skipped unless the
module knows FOO, so no changes are possible w/o a good classname.
Without class FOO in a query/command, only ids=31 can be changed by
that query. This protects all class'd prdbgs from changes by legacy
class-less user queries.
With this support, the module opt-in approach means that:
- modules declare classnames they like, meaningful names: DRM_UT_*
these are numbered [0..N]
- modules call pr_debug_class(N, "fmt..",...)
or drm_dbg(CAT, "fmt..",...) - same form.
- class-id space, while limited:0-30, is private to each module
- "class FOO" is only way to select a class'd prdbg
- unrelated modules use 0..N separately, for different purposes.
- modules "share" classnames by separate decls (uses of macro)
all drm modules reuse the above declaration.
then they respond together to a >control
4 CLASS_TYPES are defined; they split behavior on 2 factors:
1. independent bits vs related:(X<Y) bits [0-N]
2. input form: number/hex vs CSV-list: [+-]CLASS_NAMES
DD_CLASS_TYPE_DISJOINT_BITS independent bits, 0x1f input is clearest
DD_CLASS_TYPE_LEVEL_NUM (x<y), input is level, not bitvec
DD_CLASS_TYPE_DISJOINT_NAMES csv-list of independent +-CLASS_NAMEs
DD_CLASS_TYPE_LEVEL_NAMES csv-list of related +-CLASS_NAMEs, only 1 term needed
Splitting up behavior this way is safe, and flexible:
- no bitmap-vs-level > sysknob ambiguity
as was case when both were accepted on same knob
- narrower interfaces
uint is uint
- can defer DISJOINT_NAMES handling, but keep the enums.
it has no users ...
- can later add 2 more ENUMS allowing both inputs
in separate LEVEL_NUM & DISJOINT_BITS choices
then authors choice if they want to accept mixed input
- can enumerate "wierd" relations if needed
DISJOINT_BITS|LEVEL_NUM should cover everything I can forsee
but theres room for DD_CLASS_TYPE_STOCHASTIC (over the garage)
DISJOINT & LEVEL cover the space; 1st is independent, 2nd is related.
The relation could differ from (x<y), but atm I can't think of another
that makes any sensible distinction from (x<y).
Quick review of debug logging:
DRM is biggest use-case: 23 macros, issuing 10 exclusive categories of
messages, from ~5100 callsites, across core (drm, _*helpers), and 3+
drivers (Im working amdgpu, i915, more recently nouveau). These mostly
flow to drm_debug_enabled(category) for a print/no decision.
nouveau has minimum direct use of the DRM.debug macros, but does use
pr_debug (in VMM iirc), and drm_debug_enabled() in ~50 places via
macros, and another ~50 iirc due to NOUVEAU_DEBUG_MMU=y
What nouveau has most of is log-levels: there are ~800 uses of them;
nvkm_(fatal|error|warn|info|debug|trace|spam), they all flow thru
nvkm_printk_. Even excluding some levels as "always-on" and not
appropriate for dyndbg's extra resources and complexity, theres ~450
callsites.
A reported smattering of raw/adhoc debug code. all but unknown to me.
Like any debug logging, they're mostly off, and should have close to
0-off-cost. This is what dyndbg (with/on jump-label) was made for.
Bottom line: in order to unify more potential (drm,etc) use-cases,
dyndbg needs to support both debug classification schemes:
RELATED/LEVEL_NUM and INDEPENDENT/DISJOINT_BITS.
This is rev3. rev2 is here:
https://lore.kernel.org/lkml/20220516225640.3102269-1-jim.cromie@gmail.com/
summary of diffs:
- rebased on 5.19-rc6 to pick up kernel/module changes
- tracfs bits now use __vstring, __vstr_assign, from S.Rostedt
- 4 class-map-types - as outlined above
now supports LEVEL_NUM semantics, WIP nouveau integration.
v2 became the DISJOINT_BITS use case
Lots of API-ish surface area here *RFC*
- class-maps now in section("__dyndbg_classes")
class FOO queries are available at earlyboot / module-load
drop (un)?register_classes()
- test-dynamic-debug module
tests the 4 CLASS-TYPES
good place to bikeshed / paintshop the API
- nouveau - start poking - WIP
NV_PRINT -> dev_dbg (creates 632 prdbgs, but not class'd)
LEVEL_NUM classes declared to see how they "fit", unused yet.
Summary:
- plenty of new stuff here.
- plenty of new API surface area.
- exposes CLASSNAMES to user, via control, >symbolic-sysknob
More on LEVEL_NUM:
dyndbg's class scheme is names-based, so even for LEVEL_NUM with
numeric input, there must be a "V0" name; a name with which to turn
off "V1". And don't do 'pr_dbg_class(0,..)'.
LEVEL_NUM is an overlay on DISJOINT_BITS; the callback reads the input as a
level, and computes and applies the bitmap to enforce it. Yes it
wastes "V0" bit. That might matter if an author wanted many separate
verbose-knobs with a few levels each, but that seems fairly unlikely.
More importantly, it means that relative verbosity is not enforced at
the >control input; IOW this is allowed:
:#> echo class V1 -p > /proc/dynamic_debug/control
:#> echo class V3 +p > /proc/dynamic_debug/control
2nd line doesn't undo 1st, as it would if changed via sysfs-knob:
:#> echo 3 > /sys/module/drm/parameters/debug
ISTM this is fine, the "relative" meaning is added at the interface,
where it presents to a user. Doing otherwise means surgery to
ddebug_change, which seems unnecessary given an option.
CLASS NAMESPACE and DRM
The class namespace is flat and wide, so "V1" is a poor choice, other
than for demonstration (as in test_dynamic_debug module). If every
module had a "V1", turning them all on would issue lots of unrelated
noise.
Instead, class DRM_UT_CORE has a name with a specific sub-system wide
(ie multi-module) meaning, and can (and should) be responded to by all
cooperating modules.
For DRM, this means 4-5 duplicate uses of DEFINE_DYNDBG_CLASSMAP, from
multiple drivers' "*_{drm,dev,main}.c" files; all of them agreeing on
the classnames, map-type and base. Though repetitive, and a bit
fiddly when multiple classes share the module's id-space, it is
simple, and consistent with the "static struct foo = {INIT_expr}"
pattern used in the macro. Im open to simpler ways, initializing an
extern sounded like a maze w/o cheese.
For core, drm_print.c has an additional task; it defines and
initializes the bitmap sysknob that converts bits to a query:
static struct ddebug_classes_bitmap_param drm_debug_bitmap = {
.bits = &__drm_debug,
.flags = "p",
.map = &drm_debug_classes,
};
module_param_cb(debug, ¶m_ops_dyndbg_classes,
&drm_debug_bitmap, 0600);
then the callback converts bit-toggles on input into something like:
:#> echo class DRM_UT_KMS +p > control
which enables all drm_dbg_kms() callsites in all modules that declared
that classname as known.
NB: classmaps and classes_bitmap_params are different structs. this
allows the former to be shared by 2 of latter, each toggling 'p' & 'T'
flags, to control output to syslog or tracelog independently.
NBB: while "V1" isnt a good classname, "EMERG" and and other
KERN-LEVEL names have a legacy. That said, dyndbg shouldnt be
involved in EMERG actions.
NOUVEAU and multiple LEVEL_NUM knobs ?
One thing nouveau has that I havent yet grokked is 2|3 independent
verbose levels, in {,cli,sub}->debug (see subdev.h). But Ive coded
them as 2 more macro invocations; using bases 10, 15 to avoid/respect
the 0-9 mapping of DRM.debug categories to those class-ids.
With this, the drm subsystem uses 20/31 of its class-id range, perhaps
63 is better, despite the cumbersome 0x-64bit interface. Note that
the sysfs-knobs do not expose the base/offset (its an inward-facing
.class_id offset); each sysfs-knob gets its own 0..N bitmap, the base
maps it to the 0..30 subrange allocated (manually, via the base arg to
macro) for this class-map.
anyway, heres how `modprobe nouveau` goes with dynamic_debug.verbose=3
[86512.693954] dyndbg: class[0]: module:nouveau base:15 len:5 ty:1
[86512.693958] dyndbg: 15: 0 NV_SUBDEV_DBG_OFF # aka V0
[86512.693959] dyndbg: 16: 1 NV_SUBDEV_DBG_INFO
[86512.693960] dyndbg: 17: 2 NV_SUBDEV_DBG_DEBUG
[86512.693960] dyndbg: 18: 3 NV_SUBDEV_DBG_TRACE
[86512.693960] dyndbg: 19: 4 NV_SUBDEV_DBG_SPAM
[86512.693961] dyndbg: class[1]: module:nouveau base:10 len:5 ty:1
[86512.693961] dyndbg: 10: 0 NV_CLI_DBG_OFF
[86512.693962] dyndbg: 11: 1 NV_CLI_DBG_INFO
[86512.693962] dyndbg: 12: 2 NV_CLI_DBG_DEBUG
[86512.693963] dyndbg: 13: 3 NV_CLI_DBG_TRACE
[86512.693963] dyndbg: 14: 4 NV_CLI_DBG_SPAM
[86512.693963] dyndbg: class[2]: module:nouveau base:0 len:10 ty:0
[86512.693964] dyndbg: 0: 0 DRM_UT_CORE
[86512.693964] dyndbg: 1: 1 DRM_UT_DRIVER
[86512.693965] dyndbg: 2: 2 DRM_UT_KMS
[86512.693965] dyndbg: 3: 3 DRM_UT_PRIME
[86512.693965] dyndbg: 4: 4 DRM_UT_ATOMIC
[86512.693966] dyndbg: 5: 5 DRM_UT_VBL
[86512.693966] dyndbg: 6: 6 DRM_UT_STATE
[86512.693966] dyndbg: 7: 7 DRM_UT_LEASE
[86512.693967] dyndbg: 8: 8 DRM_UT_DP
[86512.693967] dyndbg: 9: 9 DRM_UT_DRMRES
[86512.693968] dyndbg: module:nouveau attached 3 classes
[86512.693968] dyndbg: 632 debug prints in module nouveau
They are not yet wired by class; the nv-printks are using dev_dbg for
now, so they can be enabled with standard selections (module, func,
etc).
DYNDBG API
I added __pr_debug_cls to make things testable, I wanted to keep it
private/un-api, because drm already has drm_dbg_<cat>(cat, fmt..), and
that seemed a pretty good model to follow: enums with strong names for
class_id constants, and same names for classes.
ie: __pr_debug_cls(DRM_UT_KMS, "Whats the Matter with KMS: %s\n", reason);
But now nouveau probably wants/needs it, perhaps others too.
pr_dbg_cls / dev_dbg_cls ? class or debug spelled out ?
MORE TESTING with DRM & TRACEFS
Most dev & testing has been on virtme, with test-dynamic-debug module
doing an api-validation-demonstration. The test module proves
multiple sysknobs, but cannot prove real multi-module sharing of the
class-name space. The DRM.debug adaptation must (and does) do that.
Now Im running on real gpu/drm hw; on both an amdgpu desktop, and this
amdgpu+nouveau laptop. Its stable enough to type on, browse, build
more kernels, etc.
:#> uname -r
5.18.0-f2-00038-g61d300083196
:#> wc /proc/dynamic_debug/control
11554 95556 1442827 /proc/dynamic_debug/control
:#> lsmod | grep drm
drm_buddy 20480 1 i915
drm_ttm_helper 16384 2 amdgpu,nouveau
ttm 86016 4 amdgpu,drm_ttm_helper,i915,nouveau
drm_dp_helper 184320 3 amdgpu,i915,nouveau
drm_kms_helper 208896 6 drm_dp_helper,amdgpu,i915,nouveau
drm 655360 16 gpu_sched,drm_dp_helper,drm_kms_helper,drm_buddy,amdgpu,drm_ttm_helper,i915,ttm,nouveau
cec 49152 2 drm_dp_helper,i915
I've never turned on DRM.debug in frustration (or curiousity until
recently); I dunno what normal traffic looks like. So I wrote a
script to toggle each category: on/off for 1 sec, so I could 'more'
thru the output to see the partitioned streams.
Different runs (of '10' sec of active tracing) produce between 330-3500
lines of logging, depending upon mousing/screen activity. Heres a bit
of a many-window-select run: `wc trace` gives "size" of the category's stream.
:#> trc_classes
12 /sys/kernel/tracing/trace
doing class DRM_UT_CORE +/- T 1394 13941 141614 /sys/kernel/tracing/trace
doing class DRM_UT_DRIVER +/- T 1396 13959 141816 /sys/kernel/tracing/trace
doing class DRM_UT_KMS +/- T 1482 14521 148246 /sys/kernel/tracing/trace
doing class DRM_UT_PRIME +/- T 1484 14539 148446 /sys/kernel/tracing/trace
doing class DRM_UT_ATOMIC +/- T 2984 31658 396064 /sys/kernel/tracing/trace
doing class DRM_UT_VBL +/- T 3411 37321 449848 /sys/kernel/tracing/trace
doing class DRM_UT_STATE +/- T 3413 37339 450048 /sys/kernel/tracing/trace
doing class DRM_UT_LEASE +/- T 3415 37357 450248 /sys/kernel/tracing/trace
doing class DRM_UT_DP +/- T 3417 37375 450442 /sys/kernel/tracing/trace
doing class DRM_UT_DRMRES +/- T 3419 37393 450644 /sys/kernel/tracing/trace
and some mid-run output from a quiet run, with less repetition.
bash-96790 [006] ..... 410237.432255: tracing_mark_write: doing class DRM_UT_DRIVER +T
bash-96790 [006] ..... 410238.610319: tracing_mark_write: doing class DRM_UT_DRIVER -T
bash-96790 [006] ..... 410239.776285: tracing_mark_write: doing class DRM_UT_KMS +T
gnome-shell-2101 [003] ..... 410239.923029: drm_debug: [FB:98]
kworker/u32:10-367584 [010] ..... 410239.924627: drm_debug: Allow idle optimizations (MALL): 0
kworker/u32:0-364714 [008] ..... 410239.935126: drm_debug: Allow idle optimizations (MALL): 1
gnome-shell-2101 [003] ..... 410240.527186: drm_debug: [FB:100]
kworker/u32:0-364714 [008] ..... 410240.528686: drm_debug: Allow idle optimizations (MALL): 0
kworker/u32:10-367584 [010] ..... 410240.539378: drm_debug: Allow idle optimizations (MALL): 1
bash-96790 [006] ..... 410240.957921: tracing_mark_write: doing class DRM_UT_KMS -T
bash-96790 [006] ..... 410242.199120: tracing_mark_write: doing class DRM_UT_PRIME +T
bash-96790 [006] ..... 410243.301988: tracing_mark_write: doing class DRM_UT_PRIME -T
bash-96790 [006] ..... 410244.416400: tracing_mark_write: doing class DRM_UT_ATOMIC +T
gnome-shell-2101 [003] ..... 410244.653372: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Allocated atomic state 000000001f1e6cb6
gnome-shell-2101 [003] ..... 410244.653381: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Added [PLANE:55:plane-3] 00000000785ae904 state to 000000001f1e6cb6
gnome-shell-2101 [003] ..... 410244.653384: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Added [CRTC:67:crtc-0] 00000000b06ca618 state to 000000001f1e6cb6
gnome-shell-2101 [003] ..... 410244.653389: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Set [FB:98] for [PLANE:55:plane-3] state 00000000785ae904
gnome-shell-2101 [003] ..... 410244.653392: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] checking 000000001f1e6cb6
gnome-shell-2101 [003] ..... 410244.653407: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] committing 000000001f1e6cb6 nonblocking
kworker/u32:0-364714 [010] ..... 410244.653729: drm_debug: amdgpu_crtc id:0 crtc_state_flags: enable:1, active:1, planes_changed:1, mode_changed:0,active_changed:0,connectors_changed:0
kworker/u32:0-364714 [010] ..... 410244.654642: drm_debug: plane: id=3 dcc_en=0
a few things to note in the trace output:
. the event-name: drm_(debug|devdbg) is not very informative
. drm_devdbg fmt has cat:%d, it could be the classname, but that should
replace event-name, not supplement it. this pertains to drm_debug too.
. drm_devdbg has lots of (dev-info), that could be optional on a new flag(s)
If anyone wants to repeat the above with the patchset, the .rc file follows:
events_3() {
local a=$1
local b=$2
local c=$3
echo $a > /sys/kernel/tracing/events/dyndbg/enable
# avoid drm_vblank*, its noisy
echo $b > /sys/kernel/tracing/events/drm/drm_debug/enable
echo $c > /sys/kernel/tracing/events/drm/drm_devdbg/enable
}
flags_class() {
local flgs=$1;
local cls=$2;
echo class $cls $flgs > /proc/dynamic_debug/control
}
# turn on each DRM_UT_* for 1 second each.
# kinda shows relative traffic of each category
# separated into blocks to get flavor of each
drm_classes() {
local flgs=$1
[ -z $flgs ] && flgs="p"
#echo > /sys/kernel/tracing/trace
#wc /sys/kernel/tracing/trace
for cls in DRM_UT_CORE DRM_UT_DRIVER DRM_UT_KMS DRM_UT_PRIME DRM_UT_ATOMIC \
DRM_UT_VBL DRM_UT_STATE DRM_UT_LEASE DRM_UT_DP \
DRM_UT_DRMRES;
do
echo -n doing class $cls " "
echo doing class $cls +$flgs
sleep .1
flags_class +$flgs $cls
sleep 1
echo doing class $cls -$flgs
sleep .1
flags_class -$flgs $cls
sleep 1
#wc /sys/kernel/tracing/trace
done
}
# like above but with T flag, and markers, trace/enable
trc_classes() {
local flgs=$1
[ -z $flgs ] && flgs="T"
echo > /sys/kernel/tracing/trace
wc -l /sys/kernel/tracing/trace
events_3 1 1 1
for cls in DRM_UT_CORE DRM_UT_DRIVER DRM_UT_KMS DRM_UT_PRIME \
DRM_UT_ATOMIC \
DRM_UT_VBL DRM_UT_STATE DRM_UT_LEASE DRM_UT_DP \
DRM_UT_DRMRES;
do
echo -n doing class $cls +/- $flgs " "
echo doing class $cls +$flgs > /sys/kernel/tracing/trace_marker
sleep .1
flags_class +$flgs $cls
sleep 1
echo doing class $cls -$flgs > /sys/kernel/tracing/trace_marker
sleep .1
flags_class -$flgs $cls
sleep 1
wc /sys/kernel/tracing/trace
done
events_3 0 0 0
}
Jim Cromie (33):
1st, a group of "cleanups"
dyndbg: fix static_branch manipulation
dyndbg: fix module.dyndbg handling
dyndbg: show both old and new in change-info
dyndbg: reverse module walk in cat control
dyndbg: reverse module.callsite walk in cat control
dyndbg: use ESCAPE_SPACE for cat control
dyndbg: let query-modname override actual module name
dyndbg: add test_dynamic_debug module
dyndbg: drop EXPORTed dynamic_debug_exec_queries
dyndbg: cleanup local vars in ddebug_init
dyndbg: create and use struct _ddebug_info
class FOO support
dyndbg: add class_id to pr_debug callsites
dyndbg: add __pr_debug_cls for testing
dyndbg: add DECLARE_DYNDBG_CLASSMAP macro
kernel/module: add __dyndbg_classes section
dyndbg: add ddebug_attach_module_classes
dyndbg: validate class FOO by checking with module
doc-dyndbg: describe "class CLASS_NAME" query support
doc-dyndbg: edit dynamic-debug-howto for brevity, audience
add dyndbg-class-param support
dyndbg: add drm.debug style (drm/parameters/debug) bitmap support
dyndbg: test DECLARE_DYNDBG_CLASSMAP, sysfs nodes
drm.debug adaptation
drm_print: condense enum drm_debug_category
drm: POC drm on dyndbg - use in core, 2 helpers, 3 drivers.
drm_print: interpose drm_*dbg with forwarding macros
drm_print: wrap drm_*_dbg in dyndbg descriptor factory macro
drm-print.h: include dyndbg header
drm-print: add drm_dbg_driver to improve namespace symmetry
drm_print: refine drm_debug_enabled for jump-label
drm_print: prefer bare printk KERN_DEBUG on generic fn
drm_print: add _ddebug descriptor to drm_*dbg prototypes
nouveau-LEVEL_NUM integration: WIP/exploratory.
nouveau: change nvkm_debug/trace to use dev_dbg POC
nouveau: adapt NV_DEBUG, NV_ATOMIC to use DRM.debug
nouveau: WIP add 2 LEVEL_NUM classmaps for CLI, SUBDEV
.../admin-guide/dynamic-debug-howto.rst | 246 +++++-----
MAINTAINERS | 2 +
drivers/gpu/drm/Kconfig | 12 +
drivers/gpu/drm/Makefile | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 14 +
drivers/gpu/drm/display/drm_dp_helper.c | 13 +
drivers/gpu/drm/drm_crtc_helper.c | 13 +
drivers/gpu/drm/drm_print.c | 48 +-
drivers/gpu/drm/i915/i915_params.c | 12 +
.../gpu/drm/nouveau/include/nvkm/core/debug.h | 16 +
.../drm/nouveau/include/nvkm/core/subdev.h | 17 +-
drivers/gpu/drm/nouveau/nouveau_drm.c | 20 +
drivers/gpu/drm/nouveau/nouveau_drv.h | 16 +-
drivers/gpu/drm/nouveau/nvkm/core/subdev.c | 23 +
include/asm-generic/vmlinux.lds.h | 3 +
include/drm/drm_print.h | 85 +++-
include/linux/dynamic_debug.h | 176 +++++--
kernel/module/internal.h | 4 +-
kernel/module/main.c | 20 +-
lib/Kconfig.debug | 10 +
lib/Makefile | 1 +
lib/dynamic_debug.c | 450 +++++++++++++++---
lib/test_dynamic_debug.c | 165 +++++++
23 files changed, 1099 insertions(+), 269 deletions(-)
create mode 100644 lib/test_dynamic_debug.c
--
2.37.1
[Catching up on mails]
Did you see the feedback form Steven from tracing side of things? Does
this version have that taken into account already?
I think once we have that and acks from dyndbg maintainers then I think we
should just land this and polish from there? Maybe some acks from drm
drivrer folks (cros people where once interested), but otherwise I'm a bit
afraid this might be stuck in limbo too much for too long and we just need
to get going :-/
-Daniel
On Fri, Aug 05, 2022 at 03:53:22PM -0600, Jim Cromie wrote:
> Hi Jason, Greg, DRM-folk,
>
> heres V5. Diffs vs V4:
>
> 1st 9 have Jason's Ack
> 10 is new: simple var cleanup
> 11 new struct, to contain future addins. touches kernel/module/
>
> dyndbg-param callbacks moved to last (of dyndbg patches)
> they're where the uncertainty is,
> and the bulk of the API addition (class keyword is small)
>
> patches between the move are the guts of the "class foo" feature.
> exposure of feature is only via >control, until sysfs parts added
>
> DRM chunk:
> included here to see how patchwork's merge & test proceeds.
> categories -> classes re-work is unchanged - feels pretty solid.
> nouveau & LEVEL made some progress, still WIP.
>
> I left out tracefs bits for now.
>
> UNCERTAINTIES / RFC:
>
> 1st is my "sharing" technique, across DRM's multi modules. Its
> literally duplicated decls, by each participating module, each adding
> static data in its own special section. I didn't try at all to shrink
> or unify this any further. There might be a perfect place to add a
> single decl, but that quest will require a pony.
>
> Sharing by class-names lacks the compiler support that varnames get;
> "incomplete" enables will probably happen if one of the declarations
> miss-spells one of the classnames.
>
> That said, see the how-to below: stringifying va-args in the macro
> would allow direct use of the enum-consts, and cc will insist that
> those consts are defined, eliminating part of the un-support.
>
> I could split the _BITS,_NUM handling out from _NAMES (ie split
> drm.debug style patch); theres no user-demand for the _NAMES feature
> ATM, and theres probably some legit API-bikeshedding to do.
>
> PATCHSET DESCRIPTION
>
> This patchset adds 'typed' "class FOO" support to dynamic-debug, where
> 'typed' means either DISJOINT_BITS, like drm_dbg(DRM_UT_CORE,..), or
> LEVEL_NUM (like nouveau debug-levels). Use it in DRM modules: core,
> helpers, and in drivers i915, amdgpu, nouveau.
>
> If a module is using class'd prdbgs (pr_debug_class, dev_dbg_class, or
> adapted drm_dbg_<category>) or similar in its code, it can "opt in" to
> allow dyndbg to manipulate those class'd prdebugs, by declaring in a
> c-file:
>
> DECLARE_DYNDBG_CLASSMAP(drm_debug_classes,
> DD_CLASS_TYPE_DISJOINT_BITS, 0,
> "DRM_UT_CORE",
> "DRM_UT_DRIVER",
> "DRM_UT_KMS",
> "DRM_UT_PRIME",
> "DRM_UT_ATOMIC",
> "DRM_UT_VBL",
> "DRM_UT_STATE",
> "DRM_UT_LEASE",
> "DRM_UT_DP",
> "DRM_UT_DRMRES");
> // how-to stringify __va_args inside the macro ?
>
> By doing this, a module tells dyndbg that it:
>
> - is using class-ids [0..N] in prdbg callsites
> ( 0..N == DRM_UT_CORE..DRM_UT_DRMRES )
> - wants to refer to them by class-names [0..N]
> - is mapping those names to those class-ids
> - expects users to enable them via >control or >parameter/knob
>
> Then, a user can enable the prdbgs by their class:
>
> :#> echo class DRM_UT_KMS +p > /proc/dynamic_debug/control
>
> And with another 3-line bitmap param decl/init, wrapping the
> drm_debug_classes var in a module-param-cb:
>
> :#> echo 0x1 > /sys/module/drm/parameters/debug
>
> and optionally using classnames:
>
> :#> echo +DRM_UT_CORE,-DRM_UT_KMS \
> > /sys/module/drm/parameters/debug_cats
>
> DYNAMIC_DEBUG gets:
>
> struct _ddebug gets a new .class_id:5 field, big enough to represent
> drm_debug_category (after squeezing). It defaults to 31 for all
> existing prdbgs. class_id also supports verbose uses.
>
> classmaps (as declared by macro above) are in their own linker
> section, are loaded by kernel/module, and handled by add_module,
> which attaches classmaps to their module's ddebug table.
>
> ddebug_change() handles a class FOO query by validating that FOO is
> known by each module in the loop. The query is skipped unless the
> module knows FOO, so no changes are possible w/o a good classname.
>
> Without class FOO in a query/command, only ids=31 can be changed by
> that query. This protects all class'd prdbgs from changes by legacy
> class-less user queries.
>
> With this support, the module opt-in approach means that:
>
> - modules declare classnames they like, meaningful names: DRM_UT_*
> these are numbered [0..N]
> - modules call pr_debug_class(N, "fmt..",...)
> or drm_dbg(CAT, "fmt..",...) - same form.
> - class-id space, while limited:0-30, is private to each module
> - "class FOO" is only way to select a class'd prdbg
> - unrelated modules use 0..N separately, for different purposes.
> - modules "share" classnames by separate decls (uses of macro)
> all drm modules reuse the above declaration.
> then they respond together to a >control
>
> 4 CLASS_TYPES are defined; they split behavior on 2 factors:
>
> 1. independent bits vs related:(X<Y) bits [0-N]
> 2. input form: number/hex vs CSV-list: [+-]CLASS_NAMES
>
> DD_CLASS_TYPE_DISJOINT_BITS independent bits, 0x1f input is clearest
> DD_CLASS_TYPE_LEVEL_NUM (x<y), input is level, not bitvec
>
> DD_CLASS_TYPE_DISJOINT_NAMES csv-list of independent +-CLASS_NAMEs
> DD_CLASS_TYPE_LEVEL_NAMES csv-list of related +-CLASS_NAMEs, only 1 term needed
>
> Splitting up behavior this way is safe, and flexible:
>
> - no bitmap-vs-level > sysknob ambiguity
> as was case when both were accepted on same knob
> - narrower interfaces
> uint is uint
> - can defer DISJOINT_NAMES handling, but keep the enums.
> it has no users ...
> - can later add 2 more ENUMS allowing both inputs
> in separate LEVEL_NUM & DISJOINT_BITS choices
> then authors choice if they want to accept mixed input
> - can enumerate "wierd" relations if needed
> DISJOINT_BITS|LEVEL_NUM should cover everything I can forsee
> but theres room for DD_CLASS_TYPE_STOCHASTIC (over the garage)
>
> DISJOINT & LEVEL cover the space; 1st is independent, 2nd is related.
> The relation could differ from (x<y), but atm I can't think of another
> that makes any sensible distinction from (x<y).
>
> Quick review of debug logging:
>
> DRM is biggest use-case: 23 macros, issuing 10 exclusive categories of
> messages, from ~5100 callsites, across core (drm, _*helpers), and 3+
> drivers (Im working amdgpu, i915, more recently nouveau). These mostly
> flow to drm_debug_enabled(category) for a print/no decision.
>
> nouveau has minimum direct use of the DRM.debug macros, but does use
> pr_debug (in VMM iirc), and drm_debug_enabled() in ~50 places via
> macros, and another ~50 iirc due to NOUVEAU_DEBUG_MMU=y
>
> What nouveau has most of is log-levels: there are ~800 uses of them;
> nvkm_(fatal|error|warn|info|debug|trace|spam), they all flow thru
> nvkm_printk_. Even excluding some levels as "always-on" and not
> appropriate for dyndbg's extra resources and complexity, theres ~450
> callsites.
>
> A reported smattering of raw/adhoc debug code. all but unknown to me.
>
> Like any debug logging, they're mostly off, and should have close to
> 0-off-cost. This is what dyndbg (with/on jump-label) was made for.
>
> Bottom line: in order to unify more potential (drm,etc) use-cases,
> dyndbg needs to support both debug classification schemes:
> RELATED/LEVEL_NUM and INDEPENDENT/DISJOINT_BITS.
>
> This is rev3. rev2 is here:
> https://lore.kernel.org/lkml/20220516225640.3102269-1-jim.cromie@gmail.com/
>
> summary of diffs:
>
> - rebased on 5.19-rc6 to pick up kernel/module changes
> - tracfs bits now use __vstring, __vstr_assign, from S.Rostedt
>
> - 4 class-map-types - as outlined above
> now supports LEVEL_NUM semantics, WIP nouveau integration.
> v2 became the DISJOINT_BITS use case
> Lots of API-ish surface area here *RFC*
>
> - class-maps now in section("__dyndbg_classes")
> class FOO queries are available at earlyboot / module-load
> drop (un)?register_classes()
>
> - test-dynamic-debug module
> tests the 4 CLASS-TYPES
> good place to bikeshed / paintshop the API
>
> - nouveau - start poking - WIP
> NV_PRINT -> dev_dbg (creates 632 prdbgs, but not class'd)
> LEVEL_NUM classes declared to see how they "fit", unused yet.
>
> Summary:
>
> - plenty of new stuff here.
> - plenty of new API surface area.
> - exposes CLASSNAMES to user, via control, >symbolic-sysknob
>
> More on LEVEL_NUM:
>
> dyndbg's class scheme is names-based, so even for LEVEL_NUM with
> numeric input, there must be a "V0" name; a name with which to turn
> off "V1". And don't do 'pr_dbg_class(0,..)'.
>
> LEVEL_NUM is an overlay on DISJOINT_BITS; the callback reads the input as a
> level, and computes and applies the bitmap to enforce it. Yes it
> wastes "V0" bit. That might matter if an author wanted many separate
> verbose-knobs with a few levels each, but that seems fairly unlikely.
>
> More importantly, it means that relative verbosity is not enforced at
> the >control input; IOW this is allowed:
>
> :#> echo class V1 -p > /proc/dynamic_debug/control
> :#> echo class V3 +p > /proc/dynamic_debug/control
>
> 2nd line doesn't undo 1st, as it would if changed via sysfs-knob:
>
> :#> echo 3 > /sys/module/drm/parameters/debug
>
> ISTM this is fine, the "relative" meaning is added at the interface,
> where it presents to a user. Doing otherwise means surgery to
> ddebug_change, which seems unnecessary given an option.
>
> CLASS NAMESPACE and DRM
>
> The class namespace is flat and wide, so "V1" is a poor choice, other
> than for demonstration (as in test_dynamic_debug module). If every
> module had a "V1", turning them all on would issue lots of unrelated
> noise.
>
> Instead, class DRM_UT_CORE has a name with a specific sub-system wide
> (ie multi-module) meaning, and can (and should) be responded to by all
> cooperating modules.
>
> For DRM, this means 4-5 duplicate uses of DEFINE_DYNDBG_CLASSMAP, from
> multiple drivers' "*_{drm,dev,main}.c" files; all of them agreeing on
> the classnames, map-type and base. Though repetitive, and a bit
> fiddly when multiple classes share the module's id-space, it is
> simple, and consistent with the "static struct foo = {INIT_expr}"
> pattern used in the macro. Im open to simpler ways, initializing an
> extern sounded like a maze w/o cheese.
>
> For core, drm_print.c has an additional task; it defines and
> initializes the bitmap sysknob that converts bits to a query:
>
> static struct ddebug_classes_bitmap_param drm_debug_bitmap = {
> .bits = &__drm_debug,
> .flags = "p",
> .map = &drm_debug_classes,
> };
> module_param_cb(debug, ¶m_ops_dyndbg_classes,
> &drm_debug_bitmap, 0600);
>
> then the callback converts bit-toggles on input into something like:
>
> :#> echo class DRM_UT_KMS +p > control
>
> which enables all drm_dbg_kms() callsites in all modules that declared
> that classname as known.
>
> NB: classmaps and classes_bitmap_params are different structs. this
> allows the former to be shared by 2 of latter, each toggling 'p' & 'T'
> flags, to control output to syslog or tracelog independently.
>
> NBB: while "V1" isnt a good classname, "EMERG" and and other
> KERN-LEVEL names have a legacy. That said, dyndbg shouldnt be
> involved in EMERG actions.
>
> NOUVEAU and multiple LEVEL_NUM knobs ?
>
> One thing nouveau has that I havent yet grokked is 2|3 independent
> verbose levels, in {,cli,sub}->debug (see subdev.h). But Ive coded
> them as 2 more macro invocations; using bases 10, 15 to avoid/respect
> the 0-9 mapping of DRM.debug categories to those class-ids.
>
> With this, the drm subsystem uses 20/31 of its class-id range, perhaps
> 63 is better, despite the cumbersome 0x-64bit interface. Note that
> the sysfs-knobs do not expose the base/offset (its an inward-facing
> .class_id offset); each sysfs-knob gets its own 0..N bitmap, the base
> maps it to the 0..30 subrange allocated (manually, via the base arg to
> macro) for this class-map.
>
> anyway, heres how `modprobe nouveau` goes with dynamic_debug.verbose=3
>
> [86512.693954] dyndbg: class[0]: module:nouveau base:15 len:5 ty:1
> [86512.693958] dyndbg: 15: 0 NV_SUBDEV_DBG_OFF # aka V0
> [86512.693959] dyndbg: 16: 1 NV_SUBDEV_DBG_INFO
> [86512.693960] dyndbg: 17: 2 NV_SUBDEV_DBG_DEBUG
> [86512.693960] dyndbg: 18: 3 NV_SUBDEV_DBG_TRACE
> [86512.693960] dyndbg: 19: 4 NV_SUBDEV_DBG_SPAM
> [86512.693961] dyndbg: class[1]: module:nouveau base:10 len:5 ty:1
> [86512.693961] dyndbg: 10: 0 NV_CLI_DBG_OFF
> [86512.693962] dyndbg: 11: 1 NV_CLI_DBG_INFO
> [86512.693962] dyndbg: 12: 2 NV_CLI_DBG_DEBUG
> [86512.693963] dyndbg: 13: 3 NV_CLI_DBG_TRACE
> [86512.693963] dyndbg: 14: 4 NV_CLI_DBG_SPAM
> [86512.693963] dyndbg: class[2]: module:nouveau base:0 len:10 ty:0
> [86512.693964] dyndbg: 0: 0 DRM_UT_CORE
> [86512.693964] dyndbg: 1: 1 DRM_UT_DRIVER
> [86512.693965] dyndbg: 2: 2 DRM_UT_KMS
> [86512.693965] dyndbg: 3: 3 DRM_UT_PRIME
> [86512.693965] dyndbg: 4: 4 DRM_UT_ATOMIC
> [86512.693966] dyndbg: 5: 5 DRM_UT_VBL
> [86512.693966] dyndbg: 6: 6 DRM_UT_STATE
> [86512.693966] dyndbg: 7: 7 DRM_UT_LEASE
> [86512.693967] dyndbg: 8: 8 DRM_UT_DP
> [86512.693967] dyndbg: 9: 9 DRM_UT_DRMRES
> [86512.693968] dyndbg: module:nouveau attached 3 classes
> [86512.693968] dyndbg: 632 debug prints in module nouveau
>
> They are not yet wired by class; the nv-printks are using dev_dbg for
> now, so they can be enabled with standard selections (module, func,
> etc).
>
> DYNDBG API
>
> I added __pr_debug_cls to make things testable, I wanted to keep it
> private/un-api, because drm already has drm_dbg_<cat>(cat, fmt..), and
> that seemed a pretty good model to follow: enums with strong names for
> class_id constants, and same names for classes.
>
> ie: __pr_debug_cls(DRM_UT_KMS, "Whats the Matter with KMS: %s\n", reason);
>
> But now nouveau probably wants/needs it, perhaps others too.
>
> pr_dbg_cls / dev_dbg_cls ? class or debug spelled out ?
>
> MORE TESTING with DRM & TRACEFS
>
> Most dev & testing has been on virtme, with test-dynamic-debug module
> doing an api-validation-demonstration. The test module proves
> multiple sysknobs, but cannot prove real multi-module sharing of the
> class-name space. The DRM.debug adaptation must (and does) do that.
>
> Now Im running on real gpu/drm hw; on both an amdgpu desktop, and this
> amdgpu+nouveau laptop. Its stable enough to type on, browse, build
> more kernels, etc.
>
> :#> uname -r
> 5.18.0-f2-00038-g61d300083196
> :#> wc /proc/dynamic_debug/control
> 11554 95556 1442827 /proc/dynamic_debug/control
> :#> lsmod | grep drm
> drm_buddy 20480 1 i915
> drm_ttm_helper 16384 2 amdgpu,nouveau
> ttm 86016 4 amdgpu,drm_ttm_helper,i915,nouveau
> drm_dp_helper 184320 3 amdgpu,i915,nouveau
> drm_kms_helper 208896 6 drm_dp_helper,amdgpu,i915,nouveau
> drm 655360 16 gpu_sched,drm_dp_helper,drm_kms_helper,drm_buddy,amdgpu,drm_ttm_helper,i915,ttm,nouveau
> cec 49152 2 drm_dp_helper,i915
>
> I've never turned on DRM.debug in frustration (or curiousity until
> recently); I dunno what normal traffic looks like. So I wrote a
> script to toggle each category: on/off for 1 sec, so I could 'more'
> thru the output to see the partitioned streams.
>
> Different runs (of '10' sec of active tracing) produce between 330-3500
> lines of logging, depending upon mousing/screen activity. Heres a bit
> of a many-window-select run: `wc trace` gives "size" of the category's stream.
>
> :#> trc_classes
> 12 /sys/kernel/tracing/trace
> doing class DRM_UT_CORE +/- T 1394 13941 141614 /sys/kernel/tracing/trace
> doing class DRM_UT_DRIVER +/- T 1396 13959 141816 /sys/kernel/tracing/trace
> doing class DRM_UT_KMS +/- T 1482 14521 148246 /sys/kernel/tracing/trace
> doing class DRM_UT_PRIME +/- T 1484 14539 148446 /sys/kernel/tracing/trace
> doing class DRM_UT_ATOMIC +/- T 2984 31658 396064 /sys/kernel/tracing/trace
> doing class DRM_UT_VBL +/- T 3411 37321 449848 /sys/kernel/tracing/trace
> doing class DRM_UT_STATE +/- T 3413 37339 450048 /sys/kernel/tracing/trace
> doing class DRM_UT_LEASE +/- T 3415 37357 450248 /sys/kernel/tracing/trace
> doing class DRM_UT_DP +/- T 3417 37375 450442 /sys/kernel/tracing/trace
> doing class DRM_UT_DRMRES +/- T 3419 37393 450644 /sys/kernel/tracing/trace
>
> and some mid-run output from a quiet run, with less repetition.
>
> bash-96790 [006] ..... 410237.432255: tracing_mark_write: doing class DRM_UT_DRIVER +T
> bash-96790 [006] ..... 410238.610319: tracing_mark_write: doing class DRM_UT_DRIVER -T
> bash-96790 [006] ..... 410239.776285: tracing_mark_write: doing class DRM_UT_KMS +T
> gnome-shell-2101 [003] ..... 410239.923029: drm_debug: [FB:98]
> kworker/u32:10-367584 [010] ..... 410239.924627: drm_debug: Allow idle optimizations (MALL): 0
> kworker/u32:0-364714 [008] ..... 410239.935126: drm_debug: Allow idle optimizations (MALL): 1
> gnome-shell-2101 [003] ..... 410240.527186: drm_debug: [FB:100]
> kworker/u32:0-364714 [008] ..... 410240.528686: drm_debug: Allow idle optimizations (MALL): 0
> kworker/u32:10-367584 [010] ..... 410240.539378: drm_debug: Allow idle optimizations (MALL): 1
> bash-96790 [006] ..... 410240.957921: tracing_mark_write: doing class DRM_UT_KMS -T
> bash-96790 [006] ..... 410242.199120: tracing_mark_write: doing class DRM_UT_PRIME +T
> bash-96790 [006] ..... 410243.301988: tracing_mark_write: doing class DRM_UT_PRIME -T
> bash-96790 [006] ..... 410244.416400: tracing_mark_write: doing class DRM_UT_ATOMIC +T
> gnome-shell-2101 [003] ..... 410244.653372: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Allocated atomic state 000000001f1e6cb6
> gnome-shell-2101 [003] ..... 410244.653381: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Added [PLANE:55:plane-3] 00000000785ae904 state to 000000001f1e6cb6
> gnome-shell-2101 [003] ..... 410244.653384: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Added [CRTC:67:crtc-0] 00000000b06ca618 state to 000000001f1e6cb6
> gnome-shell-2101 [003] ..... 410244.653389: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] Set [FB:98] for [PLANE:55:plane-3] state 00000000785ae904
> gnome-shell-2101 [003] ..... 410244.653392: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] checking 000000001f1e6cb6
> gnome-shell-2101 [003] ..... 410244.653407: drm_devdbg: cat:4, (0xffff8f26c173de00:0000:06:00.0)[UNSAFE-MEMORY] committing 000000001f1e6cb6 nonblocking
> kworker/u32:0-364714 [010] ..... 410244.653729: drm_debug: amdgpu_crtc id:0 crtc_state_flags: enable:1, active:1, planes_changed:1, mode_changed:0,active_changed:0,connectors_changed:0
> kworker/u32:0-364714 [010] ..... 410244.654642: drm_debug: plane: id=3 dcc_en=0
>
> a few things to note in the trace output:
>
> . the event-name: drm_(debug|devdbg) is not very informative
>
> . drm_devdbg fmt has cat:%d, it could be the classname, but that should
> replace event-name, not supplement it. this pertains to drm_debug too.
>
> . drm_devdbg has lots of (dev-info), that could be optional on a new flag(s)
>
>
> If anyone wants to repeat the above with the patchset, the .rc file follows:
>
> events_3() {
> local a=$1
> local b=$2
> local c=$3
> echo $a > /sys/kernel/tracing/events/dyndbg/enable
> # avoid drm_vblank*, its noisy
> echo $b > /sys/kernel/tracing/events/drm/drm_debug/enable
> echo $c > /sys/kernel/tracing/events/drm/drm_devdbg/enable
> }
>
> flags_class() {
> local flgs=$1;
> local cls=$2;
> echo class $cls $flgs > /proc/dynamic_debug/control
> }
>
> # turn on each DRM_UT_* for 1 second each.
> # kinda shows relative traffic of each category
> # separated into blocks to get flavor of each
>
> drm_classes() {
> local flgs=$1
> [ -z $flgs ] && flgs="p"
>
> #echo > /sys/kernel/tracing/trace
> #wc /sys/kernel/tracing/trace
>
> for cls in DRM_UT_CORE DRM_UT_DRIVER DRM_UT_KMS DRM_UT_PRIME DRM_UT_ATOMIC \
> DRM_UT_VBL DRM_UT_STATE DRM_UT_LEASE DRM_UT_DP \
> DRM_UT_DRMRES;
> do
> echo -n doing class $cls " "
>
> echo doing class $cls +$flgs
> sleep .1
> flags_class +$flgs $cls
> sleep 1
>
> echo doing class $cls -$flgs
> sleep .1
> flags_class -$flgs $cls
> sleep 1
>
> #wc /sys/kernel/tracing/trace
> done
> }
>
> # like above but with T flag, and markers, trace/enable
> trc_classes() {
> local flgs=$1
> [ -z $flgs ] && flgs="T"
>
> echo > /sys/kernel/tracing/trace
> wc -l /sys/kernel/tracing/trace
>
> events_3 1 1 1
> for cls in DRM_UT_CORE DRM_UT_DRIVER DRM_UT_KMS DRM_UT_PRIME \
> DRM_UT_ATOMIC \
> DRM_UT_VBL DRM_UT_STATE DRM_UT_LEASE DRM_UT_DP \
> DRM_UT_DRMRES;
> do
> echo -n doing class $cls +/- $flgs " "
>
> echo doing class $cls +$flgs > /sys/kernel/tracing/trace_marker
> sleep .1
> flags_class +$flgs $cls
> sleep 1
>
> echo doing class $cls -$flgs > /sys/kernel/tracing/trace_marker
> sleep .1
> flags_class -$flgs $cls
> sleep 1
>
> wc /sys/kernel/tracing/trace
>
> done
> events_3 0 0 0
> }
>
>
> Jim Cromie (33):
>
> 1st, a group of "cleanups"
>
> dyndbg: fix static_branch manipulation
> dyndbg: fix module.dyndbg handling
> dyndbg: show both old and new in change-info
> dyndbg: reverse module walk in cat control
> dyndbg: reverse module.callsite walk in cat control
> dyndbg: use ESCAPE_SPACE for cat control
> dyndbg: let query-modname override actual module name
> dyndbg: add test_dynamic_debug module
> dyndbg: drop EXPORTed dynamic_debug_exec_queries
> dyndbg: cleanup local vars in ddebug_init
> dyndbg: create and use struct _ddebug_info
>
> class FOO support
>
> dyndbg: add class_id to pr_debug callsites
> dyndbg: add __pr_debug_cls for testing
> dyndbg: add DECLARE_DYNDBG_CLASSMAP macro
> kernel/module: add __dyndbg_classes section
> dyndbg: add ddebug_attach_module_classes
> dyndbg: validate class FOO by checking with module
> doc-dyndbg: describe "class CLASS_NAME" query support
> doc-dyndbg: edit dynamic-debug-howto for brevity, audience
>
> add dyndbg-class-param support
>
> dyndbg: add drm.debug style (drm/parameters/debug) bitmap support
> dyndbg: test DECLARE_DYNDBG_CLASSMAP, sysfs nodes
>
> drm.debug adaptation
>
> drm_print: condense enum drm_debug_category
> drm: POC drm on dyndbg - use in core, 2 helpers, 3 drivers.
> drm_print: interpose drm_*dbg with forwarding macros
> drm_print: wrap drm_*_dbg in dyndbg descriptor factory macro
> drm-print.h: include dyndbg header
> drm-print: add drm_dbg_driver to improve namespace symmetry
> drm_print: refine drm_debug_enabled for jump-label
> drm_print: prefer bare printk KERN_DEBUG on generic fn
> drm_print: add _ddebug descriptor to drm_*dbg prototypes
>
> nouveau-LEVEL_NUM integration: WIP/exploratory.
>
> nouveau: change nvkm_debug/trace to use dev_dbg POC
> nouveau: adapt NV_DEBUG, NV_ATOMIC to use DRM.debug
> nouveau: WIP add 2 LEVEL_NUM classmaps for CLI, SUBDEV
>
> .../admin-guide/dynamic-debug-howto.rst | 246 +++++-----
> MAINTAINERS | 2 +
> drivers/gpu/drm/Kconfig | 12 +
> drivers/gpu/drm/Makefile | 2 +
> drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 14 +
> drivers/gpu/drm/display/drm_dp_helper.c | 13 +
> drivers/gpu/drm/drm_crtc_helper.c | 13 +
> drivers/gpu/drm/drm_print.c | 48 +-
> drivers/gpu/drm/i915/i915_params.c | 12 +
> .../gpu/drm/nouveau/include/nvkm/core/debug.h | 16 +
> .../drm/nouveau/include/nvkm/core/subdev.h | 17 +-
> drivers/gpu/drm/nouveau/nouveau_drm.c | 20 +
> drivers/gpu/drm/nouveau/nouveau_drv.h | 16 +-
> drivers/gpu/drm/nouveau/nvkm/core/subdev.c | 23 +
> include/asm-generic/vmlinux.lds.h | 3 +
> include/drm/drm_print.h | 85 +++-
> include/linux/dynamic_debug.h | 176 +++++--
> kernel/module/internal.h | 4 +-
> kernel/module/main.c | 20 +-
> lib/Kconfig.debug | 10 +
> lib/Makefile | 1 +
> lib/dynamic_debug.c | 450 +++++++++++++++---
> lib/test_dynamic_debug.c | 165 +++++++
> 23 files changed, 1099 insertions(+), 269 deletions(-)
> create mode 100644 lib/test_dynamic_debug.c
>
> --
> 2.37.1
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
© 2016 - 2024 Red Hat, Inc.