Now we have:
[acme@filo pahole]$ pahole --help
Usage: pahole [OPTION...] [FILE] {[CLASS]}
-a, --anon_include include anonymous classes
-A, --nested_anon_include include nested (inside other structs) anonymous
classes
-B, --bit_holes=NR_HOLES Show only structs at least NR_HOLES bit holes
-c, --cacheline_size=SIZE set cacheline size to SIZE
-D, --decl_exclude=PREFIX exclude classes declared in files with PREFIX
-E, --expand_types expand class members
-H, --holes=NR_HOLES show only structs at least NR_HOLES holes
-m, --nr_methods show number of methods
-n, --nr_members show number of members
-N, --class_name_len show size of classes
-P, --packable show only structs that has holes that can be
packed
-R, --reorganize reorg struct trying to kill holes
-s, --sizes show size of classes
-S, --show_reorg_steps show the struct layout at each reorganization step
-t, --nr_definitions show how many times struct was defined
-V, --verbose be verbose
-x, --exclude=PREFIX exclude PREFIXed classes
-X, --cu_exclude=PREFIX exclude PREFIXed compilation units
Input Selection:
--debuginfo-path=PATH Search path for separate debuginfo files
-e, --executable=FILE Find addresses in FILE
-k, --kernel Find addresses in the running kernel
-K, --offline-kernel[=RELEASE] Kernel with all modules
-M, --linux-process-map=FILE Find addresses in files mapped as read from
FILE in Linux /proc/PID/maps format
-p, --pid=PID Find addresses in files mapped into process PID
-?, --help Give this help list
--usage Give a short usage message
Mandatory or optional arguments to long options are also mandatory or optional
for any corresponding short options.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Now there is a cus__loadfl function that receives the tool argp tables and uses
libdwfl to process the DWARF info, with this RELA objects such as .o and .ko
files in a Linux kernel build are supported, and all the other goodies that
come from using libdwfl, such as separate debuginfo files, etc come as a bonus.
Now to convert the tools, pahole being the first, that already works well using
cus__loadfl().
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
[acme@filo examples]$ pahole -a mpg_audio_frame_t
/* <14f> /home/acme/git/pahole/examples/mpg_audio_frame_t.c:4 */
typedef struct {
uint16_t frame_sync; /* 0 2 */
uint8_t layer; /* 2 1 */
/* WARNING: DWARF offset=0, real offset=3 */
uint32_t mpeg25_bit:1; /* 0 4 */
uint32_t lsf_bit:1; /* 0 4 */
uint32_t bitrate_idx:4; /* 0 4 */
<SNIP>
So gcc combined a uint16_t + a uint8_t + the first entries in the uint32_t
bitfield that could fit in the 8 bits after the first two fields but haven't
updated the size of the bitfield (4) and wrote 0 as the offset, warn about such
inconsistencies.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
So that we print the boundary after the last member in a bitfield.
Spotted by Matthew Wilcox.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
[acme@mica pahole]$ pahole lala
pahole: Permission denied
[acme@mica pahole]$ pahole foo
pahole: No such file or directory
[acme@mica pahole]$ pahole ctracer.c
pahole: couldn't load DWARF info from ctracer.c
[acme@mica pahole]$
Thanks to Matthew Wilcox for noticing how lame it was :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
And let the final binaries get it from that. Fixes building with
-Wl,--as-needed
Signed-off-by: Diego "Flameeyes" Pettenò <flameeyes@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
[acme@filo pahole]$ cat examples/expand_typedefs.c
<SNIP>
typedef struct {
int a, b, c;
} inner;
static struct outer {
int q, b;
inner m;
} foo;
<SNIP>
[acme@filo pahole]$ pahole --expand_types examples/expand_typedefs
/* <158> /home/acme/git/pahole/examples/expand_typedefs.c:7 */
struct outer {
int q; /* 0 4 */
int b; /* 4 4 */
/* typedef inner */ struct {
int a; /* 0 4 */
int b; /* 4 4 */
int c; /* 8 4 */
} m; /* 8 12 */
}; /* size: 20, cachelines: 1 */
/* last cacheline: 20 bytes */
/* definitions: 1 */
[acme@filo pahole]$
For now it does all typedef expansions, which in at least the base types may be
a bit too much, e.g. u32 -> unsigned long int, lets see if somebody complains,
perhaps even myself 8) If that is the case we can add yet another command line
option to specify that such base type expansions should be filtered out, making
the expand_types parameter be flag mask, not just a boolean as it is today.
To see a more complete output look at:
http://oops.ghostprotocols.net:81/acme/dwarves/vmlinux-pahole-expand_types-typedef_unfolding.txt
Suggested by Jeff Muizelaar.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To tell where to find the elfutils/libdw.h, etc if they are not in the standard
system include dirs, i.e. its no use to find the paths if we don't tell the
cmake system to use the paths found 8)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that we get rid of all the buffer limits, if we need to format into strings
we can use string streams, like we're doing now in just one case, tag__name for
DW_TAG_subroutine_type, that is bogus as it is, as we need to have the name of
the type inside the type declaration (void (*type_name)(parameters)) and not
after (void (*)(parameters) type_name)), but leave this for an upcoming cset.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
1. We now use relayfs
2. ostra-cg is shipped and used in the Makefile
3. README.ctracer has all the details
I followed README.ctracer and the result, in my workstation, was this one:
http://oops.ghostprotocols.net:81/acme/dwarves/callgraphs/acme_eating_his_dog_food/6.html
I've started on the 6th page as it is more interesting, having tcp and netlink
callchains, don't be disappointed by some pages having just one level of
nesting, that is just the lack of containers and aliases (to follow tcp_sock,
skb->sk, etc) from the OSTRA days, but it's getting there! :-)
Ah, it is collecting all the base types in the specified struct by doing a
struct "view", i.e. trimming the struct to have just members that are
"reducible" to basic types (int, long, char, signed and unsigned), then
reorganizing it with the code introduced in pahole --reorganize to get it to
the best layout possible, reducing the size of the entry probe points trace
entries.
Updated rpms are available at the usual place:
http://oops.ghostprotocols.net:81/acme/dwarves/rpm/
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
And use it in a new tool, pglobal, that shows global variables and functions.
Signed-off-by: Davi Arnaut <davi@haxent.com.br>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that in tools like ctracer we can print to a file, most of the tools just
pass stdout, keeping the previous behaviour.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Will be used in ctracer to create a struct subset with just the types for which
we have "collectors", i.e. functions that reduce complex types to base types
that will be put in the mini-struct, that will be as tightly packed as it can
be.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This cset also does a fixup for cases where the compiler keeps the type
specified by the programmer for a bitfield but uses less space to combine with
the next, non-bitfield member, these cases can be caught using plain pahole and
will appear with this comment:
/* --- cacheline 1 boundary (64 bytes) --- */
int bitfield1:1; /* 64 4 */
int bitfield2:1; /* 64 4 */
/* XXX 14 bits hole, try to pack */
/* Bitfield WARNING: DWARF size=4, real size=2 */
short int d; /* 66 2 */
The fixup is done prior to reorganizing the fields.
Now an example of this code in action:
[acme@filo examples]$ cat swiss_cheese.c
<SNIP>
struct cheese {
char id;
short number;
char name[52];
int a:1;
int b;
int bitfield1:1;
int bitfield2:1;
short d;
short e;
short last:5;
};
<SNIP>
[acme@filo examples]$
Lets look at the layout:
[acme@filo examples]$ pahole swiss_cheese cheese
/* <11b> /home/acme/git/pahole/examples/swiss_cheese.c:3 */
struct cheese {
char id; /* 0 1 */
/* XXX 1 byte hole, try to pack */
short int number; /* 2 2 */
char name[52]; /* 4 52 */
int a:1; /* 56 4 */
/* XXX 31 bits hole, try to pack */
int b; /* 60 4 */
/* --- cacheline 1 boundary (64 bytes) --- */
int bitfield1:1; /* 64 4 */
int bitfield2:1; /* 64 4 */
/* XXX 14 bits hole, try to pack */
/* Bitfield WARNING: DWARF size=4, real size=2 */
short int d; /* 66 2 */
short int e; /* 68 2 */
short int last:5; /* 70 2 */
}; /* size: 72, cachelines: 2 */
/* sum members: 71, holes: 1, sum holes: 1 */
/* bit holes: 2, sum bit holes: 45 bits */
/* bit_padding: 11 bits */
/* last cacheline: 8 bytes */
[acme@filo examples]$
Full of holes, has bit padding and uses more than one 64 bytes cacheline.
Now lets ask pahole to reorganize it:
[acme@filo examples]$ pahole --reorganize --verbose swiss_cheese cheese
/* Demoting bitfield ('a' ... 'a') from 'int' to 'unsigned char' */
/* Demoting bitfield ('bitfield1' ... 'bitfield2') from 'short unsigned int' to 'unsigned char' */
/* Demoting bitfield ('last') from 'short int' to 'unsigned char' */
/* Moving 'bitfield2:1' from after 'bitfield1' to after 'a:1' */
/* Moving 'bitfield1:1' from after 'b' to after 'bitfield2:1' */
/* Moving 'last:5' from after 'e' to after 'bitfield1:1' */
/* Moving bitfield('a' ... 'last') from after 'name' to after 'id' */
/* Moving 'e' from after 'd' to after 'b' */
/* <11b> /home/acme/git/pahole/examples/swiss_cheese.c:3 */
struct cheese {
char id; /* 0 1 */
unsigned char a:1; /* 1 1 */
unsigned char bitfield2:1; /* 1 1 */
unsigned char bitfield1:1; /* 1 1 */
unsigned char last:5; /* 1 1 */
short int number; /* 2 2 */
char name[52]; /* 4 52 */
int b; /* 56 4 */
short int e; /* 60 2 */
short int d; /* 62 2 */
/* --- cacheline 1 boundary (64 bytes) --- */
}; /* size: 64, cachelines: 1 */
/* saved 8 bytes and 1 cacheline! */
[acme@filo examples]$
Instant karma, it gets completely packed, and look ma, no
__attribute__((packed)) :-)
With this struct task_struct in the linux kernel is shrunk by 12 bytes, there
is more 4 bytes to save with another technique that involves not combining
holes, but using the last single hole to fill it with members at the tail of
the struct.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This allows us to save 4 more bytes in struct task_struct, for instance, now we
need to combine whole bitfields with other fields if some bitfield has a size
less than sizeof(void *) and there is a suitable hole.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Arnaldo Carvalho de Melo
Diego "Flameeyes" Pettenò
Arnaldo Carvalho de Melo
Davi Arnaut