Now the trace action and notify of blktrace may include cgroup info
in them. If __BLK_TA_CGROUP & blk_io_trace::action is true, the cgroup
info will be stored after the blk_io_trace struct (namely the header),
and the PDU will be appended after the cgroup info.
To support trace with cgroup info, we define macro BLK_AUX_TA and
function pdu_start_len(). BLK_AUX_TA is used to strip __BLK_TA_CGROUP
from blk_io_trace::action and pdu_start_len() is used to get the start
address and the length of PDU.
Signed-off-by: Hou Tao <houtao1@xxxxxxxxxx>
---
blkparse.c | 49 +++++++++++++++++++++++++++++++------------------
blkparse_fmt.c | 22 +++++++++++++++++++++-
blktrace.h | 5 +++++
blktrace_api.h | 10 ++++++++++
4 files changed, 67 insertions(+), 19 deletions(-)
diff --git a/blkparse.c b/blkparse.c
index 227cc44..fa2f520 100644
--- a/blkparse.c
+++ b/blkparse.c
@@ -574,16 +574,20 @@ static struct process_pid_map *add_ppm_hash(pid_t pid, const char *name)
static void handle_notify(struct blk_io_trace *bit)
{
- void *payload = (caddr_t) bit + sizeof(*bit);
+ void *payload;
+ __u16 pdu_len;
__u32 two32[2];
+ __u32 act = bit->action & ~BLK_AUX_TA;
- switch (bit->action) {
+ pdu_start_len(bit, &payload, &pdu_len);
+
+ switch (act) {
case BLK_TN_PROCESS:
add_ppm_hash(bit->pid, payload);
break;
case BLK_TN_TIMESTAMP:
- if (bit->pdu_len != sizeof(two32))
+ if (pdu_len != sizeof(two32))
return;
memcpy(two32, payload, sizeof(two32));
if (!data_is_native) {
@@ -601,11 +605,11 @@ static void handle_notify(struct blk_io_trace *bit)
break;
case BLK_TN_MESSAGE:
- if (bit->pdu_len > 0) {
- char msg[bit->pdu_len+1];
+ if (pdu_len > 0) {
+ char msg[pdu_len+1];
- memcpy(msg, (char *)payload, bit->pdu_len);
- msg[bit->pdu_len] = '\0';
+ memcpy(msg, (char *)payload, pdu_len);
+ msg[pdu_len] = '\0';
fprintf(ofp,
"%3d,%-3d %2d %8s %5d.%09lu %5u %2s %3s %s\n",
@@ -1491,7 +1495,7 @@ static void dump_trace_pc(struct blk_io_trace *t, struct per_dev_info *pdi,
struct per_cpu_info *pci)
{
int w = (t->action & BLK_TC_ACT(BLK_TC_WRITE)) != 0;
- int act = t->action & 0xffff;
+ int act = (t->action & 0xffff) & ~BLK_AUX_TA;
switch (act) {
case __BLK_TA_QUEUE:
@@ -1540,7 +1544,7 @@ static void dump_trace_fs(struct blk_io_trace *t, struct per_dev_info *pdi,
struct per_cpu_info *pci)
{
int w = (t->action & BLK_TC_ACT(BLK_TC_WRITE)) != 0;
- int act = t->action & 0xffff;
+ int act = (t->action & 0xffff) & ~BLK_AUX_TA;
switch (act) {
case __BLK_TA_QUEUE:
@@ -1622,10 +1626,12 @@ static void dump_trace_fs(struct blk_io_trace *t, struct per_dev_info *pdi,
static void dump_trace(struct blk_io_trace *t, struct per_cpu_info *pci,
struct per_dev_info *pdi)
{
+ __u32 act = t->action & ~BLK_AUX_TA;
+
if (text_output) {
- if (t->action == BLK_TN_MESSAGE)
+ if (act == BLK_TN_MESSAGE)
handle_notify(t);
- else if (t->action & BLK_TC_ACT(BLK_TC_PC))
+ else if (act & BLK_TC_ACT(BLK_TC_PC))
dump_trace_pc(t, pdi, pci);
else
dump_trace_fs(t, pdi, pci);
@@ -1637,8 +1643,8 @@ static void dump_trace(struct blk_io_trace *t, struct per_cpu_info *pci,
pdi->events++;
if (bin_output_msgs ||
- !(t->action & BLK_TC_ACT(BLK_TC_NOTIFY) &&
- t->action == BLK_TN_MESSAGE))
+ !(act & BLK_TC_ACT(BLK_TC_NOTIFY) &&
+ act == BLK_TN_MESSAGE))
output_binary(t, sizeof(*t) + t->pdu_len);
}
@@ -2107,6 +2113,8 @@ static void show_entries_rb(int force)
struct trace *t;
while ((n = rb_first(&rb_sort_root)) != NULL) {
+ __u32 act;
+
if (is_done() && !force && !pipeline)
break;
@@ -2127,7 +2135,8 @@ static void show_entries_rb(int force)
break;
}
- if (!(bit->action == BLK_TN_MESSAGE) &&
+ act = bit->action & ~BLK_AUX_TA;
+ if (!(act == BLK_TN_MESSAGE) &&
check_sequence(pdi, t, force))
break;
@@ -2139,12 +2148,12 @@ static void show_entries_rb(int force)
if (!pci || pci->cpu != bit->cpu)
pci = get_cpu_info(pdi, bit->cpu);
- if (!(bit->action == BLK_TN_MESSAGE))
+ if (!(act == BLK_TN_MESSAGE))
pci->last_sequence = bit->sequence;
pci->nelems++;
- if (bit->action & (act_mask << BLK_TC_SHIFT))
+ if (act & (act_mask << BLK_TC_SHIFT))
dump_trace(bit, pci, pdi);
put_trace(pdi, t);
@@ -2225,6 +2234,7 @@ static int read_events(int fd, int always_block, int *fdblock)
struct trace *t;
int pdu_len, should_block, ret;
__u32 magic;
+ __u32 act;
bit = bit_alloc();
@@ -2273,7 +2283,8 @@ static int read_events(int fd, int always_block, int *fdblock)
/*
* not a real trace, so grab and handle it here
*/
- if (bit->action & BLK_TC_ACT(BLK_TC_NOTIFY) && bit->action != BLK_TN_MESSAGE) {
+ act = bit->action & ~BLK_AUX_TA;
+ if (act & BLK_TC_ACT(BLK_TC_NOTIFY) && act != BLK_TN_MESSAGE) {
handle_notify(bit);
output_binary(bit, sizeof(*bit) + bit->pdu_len);
continue;
@@ -2375,6 +2386,7 @@ static void ms_sort(struct ms_stream *msp)
static int ms_prime(struct ms_stream *msp)
{
__u32 magic;
+ __u32 act;
unsigned int i;
struct trace *t;
struct per_dev_info *pdi = msp->pdi;
@@ -2422,7 +2434,8 @@ static int ms_prime(struct ms_stream *msp)
continue;
}
- if (bit->action & BLK_TC_ACT(BLK_TC_NOTIFY) && bit->action != BLK_TN_MESSAGE) {
+ act = bit->action & ~BLK_AUX_TA;
+ if (act & BLK_TC_ACT(BLK_TC_NOTIFY) && act != BLK_TN_MESSAGE) {
handle_notify(bit);
output_binary(bit, sizeof(*bit) + bit->pdu_len);
bit_free(bit);
diff --git a/blkparse_fmt.c b/blkparse_fmt.c
index c42e6d7..8dc20ca 100644
--- a/blkparse_fmt.c
+++ b/blkparse_fmt.c
@@ -145,7 +145,27 @@ static char *dump_pdu(unsigned char *pdu_buf, int pdu_len)
return p;
}
-#define pdu_start(t) (((void *) (t) + sizeof(struct blk_io_trace)))
+void pdu_start_len(struct blk_io_trace *bit, void **pdu, __u16 *pdu_len)
+{
+ __u16 extra = 0;
+
+ if (bit->action & __BLK_TA_CGROUP) {
+ extra += sizeof(union kernfs_node_id);
+ }
+
+ *pdu = (caddr_t)bit + sizeof(*bit) + extra;
+ if (pdu_len)
+ *pdu_len = bit->pdu_len - extra;
+}
+
+static void *pdu_start(struct blk_io_trace *t)
+{
+ void *pdu;
+
+ pdu_start_len(t, &pdu, NULL);
+
+ return pdu;
+}
static unsigned int get_pdu_int(struct blk_io_trace *t)
{
diff --git a/blktrace.h b/blktrace.h
index 944fc08..c5ed618 100644
--- a/blktrace.h
+++ b/blktrace.h
@@ -26,6 +26,8 @@
#define t_kb(t) ((t)->bytes >> 10)
#define t_b(t) ((t)->bytes & 1023)
+#define BLK_AUX_TA (__BLK_TA_CGROUP)
+
typedef __u32 u32;
typedef __u8 u8;
@@ -148,4 +150,7 @@ extern int valid_act_opt(int);
extern int find_mask_map(char *);
extern char *find_process_name(pid_t);
+extern void pdu_start_len(struct blk_io_trace *bit,
+ void **pdu, __u16 *pdu_len);
+
#endif
diff --git a/blktrace_api.h b/blktrace_api.h
index b222218..a78a862 100644
--- a/blktrace_api.h
+++ b/blktrace_api.h
@@ -51,6 +51,7 @@ enum {
__BLK_TA_REMAP, /* bio was remapped */
__BLK_TA_ABORT, /* request aborted */
__BLK_TA_DRV_DATA, /* binary driver data */
+ __BLK_TA_CGROUP = 1 << 8,
};
/*
@@ -60,6 +61,7 @@ enum blktrace_notify {
__BLK_TN_PROCESS = 0, /* establish pid/name mapping */
__BLK_TN_TIMESTAMP, /* include system clock */
__BLK_TN_MESSAGE, /* Character string message */
+ __BLK_TN_CGROUP = __BLK_TA_CGROUP,
};
/*
@@ -90,6 +92,14 @@ enum blktrace_notify {
#define BLK_IO_TRACE_MAGIC 0x65617400
#define BLK_IO_TRACE_VERSION 0x07
+union kernfs_node_id {
+ struct {
+ __u32 ino;
+ __u32 generation;
+ };
+ __u64 id;
+};
+
/*
* The trace itself
*/
--
2.9.5
--
To unsubscribe from this list: send the line "unsubscribe linux-btrace" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
