This patch allows relay channels to be reset i.e. unconsumed.
Basically allows a 'rewind' function for flight-recorder tracing.
Signed-off-by: Tom Zanussi <[email protected]>
Signed-off-by: David Wilder <[email protected]>
---
Documentation/filesystems/relay.txt | 11 ++++++
include/linux/relay.h | 1 +
kernel/relay.c | 58 ++++++++++++++++++++++++++++++++---
3 files changed, 65 insertions(+), 5 deletions(-)
diff --git a/Documentation/filesystems/relay.txt b/Documentation/filesystems/relay.txt
index 094f2d2..1af6d4b 100644
--- a/Documentation/filesystems/relay.txt
+++ b/Documentation/filesystems/relay.txt
@@ -161,6 +161,7 @@ TBD(curr. line MT:/API/)
relay_close(chan)
relay_flush(chan)
relay_reset(chan)
+ relay_reset_consumed(chan)
channel management typically called on instigation of userspace:
@@ -452,6 +453,16 @@ state without reallocating channel buffer memory or destroying
existing mappings. It should however only be called when it's safe to
do so, i.e. when the channel isn't currently being written to.
+The read(2) implementation always 'consumes' the bytes read,
+i.e. those bytes won't be available again to subsequent reads.
+Certain applications may nonetheless wish to allow the 'consumed' data
+to be re-read; relay_reset_consumed() is provided for that purpose -
+it resets the internal consumed counters for all buffers in the
+channel. For example, if a first set of reads 'drains' the channel,
+and then relay_reset_consumed() is called, a second set of reads will
+get the exact same data (assuming no new data was written between the
+first set of reads and the second).
+
Finally, there are a couple of utility callbacks that can be used for
different purposes. buf_mapped() is called whenever a channel buffer
is mmapped from user space and buf_unmapped() is called when it's
diff --git a/include/linux/relay.h b/include/linux/relay.h
index 6cd8c44..aca45fa 100644
--- a/include/linux/relay.h
+++ b/include/linux/relay.h
@@ -175,6 +175,7 @@ extern void relay_subbufs_consumed(struct rchan *chan,
unsigned int cpu,
size_t consumed);
extern void relay_reset(struct rchan *chan);
+extern void relay_reset_consumed(struct rchan *chan);
extern int relay_buf_full(struct rchan_buf *buf);
extern size_t relay_switch_subbuf(struct rchan_buf *buf,
diff --git a/kernel/relay.c b/kernel/relay.c
index d080b9d..f9dc539 100644
--- a/kernel/relay.c
+++ b/kernel/relay.c
@@ -378,6 +378,57 @@ void relay_reset(struct rchan *chan)
}
EXPORT_SYMBOL_GPL(relay_reset);
+/**
+ * __relay_reset_consumed - reset a channel buffer's consumed count
+ * @buf: the channel buffer
+ *
+ * See relay_reset_consumed for description of effect.
+ */
+static void __relay_reset_consumed(struct rchan_buf *buf)
+{
+ size_t n_subbufs = buf->chan->n_subbufs;
+ size_t produced = buf->subbufs_produced;
+ size_t consumed = buf->subbufs_consumed;
+
+ if (produced < n_subbufs)
+ buf->subbufs_consumed = 0;
+ else {
+ consumed = produced - n_subbufs;
+ if (buf->offset)
+ consumed++;
+ buf->subbufs_consumed = consumed;
+ }
+ buf->bytes_consumed = 0;
+}
+
+/**
+ * relay_reset_consumed - reset the channel's consumed counts
+ * @chan: the channel
+ *
+ * This has the effect of making all data previously read (and
+ * not overwritten by subsequent writes) from a channel available
+ * for reading again.
+ *
+ * NOTE: Care should be taken that the channel isn't actually
+ * being used by anything when this call is made.
+ */
+void relay_reset_consumed(struct rchan *chan)
+{
+ unsigned int i;
+ struct rchan_buf *prev = NULL;
+
+ if (!chan)
+ return;
+
+ for (i = 0; i < NR_CPUS; i++) {
+ if (!chan->buf[i] || chan->buf[i] == prev)
+ break;
+ __relay_reset_consumed(chan->buf[i]);
+ prev = chan->buf[i];
+ }
+}
+EXPORT_SYMBOL_GPL(relay_reset_consumed);
+
/*
* relay_open_buf - create a new relay channel buffer
*
@@ -840,11 +891,8 @@ static int relay_file_read_avail(struct rchan_buf *buf, size_t read_pos)
return 1;
}
- if (unlikely(produced - consumed >= n_subbufs)) {
- consumed = produced - n_subbufs + 1;
- buf->subbufs_consumed = consumed;
- buf->bytes_consumed = 0;
- }
+ if (unlikely(produced - consumed >= n_subbufs))
+ __relay_reset_consumed(buf);
produced = (produced % n_subbufs) * subbuf_size + buf->offset;
consumed = (consumed % n_subbufs) * subbuf_size + buf->bytes_consumed;
On Mon, 03 Mar 2008 15:51:50 -0800
"David J. Wilder" <[email protected]> wrote:
> This patch allows relay channels to be reset i.e. unconsumed.
> Basically allows a 'rewind' function for flight-recorder tracing.
>
> +void relay_reset_consumed(struct rchan *chan)
> +{
> + unsigned int i;
> + struct rchan_buf *prev = NULL;
> +
> + if (!chan)
> + return;
> +
> + for (i = 0; i < NR_CPUS; i++) {
Use of NR_CPUS is usually wrong. In this case it seems you should be using
for_each_possible_cpu().
Also the existing usage of NR_CPUS in relay_subbufs_consumed() should be
switched to using cpu_possible().
New usage of NR_CPUS might be checkpatch-worthy, actually:
akpm:/usr/src/25> grep -l '^+.*NR_CPUS' patches/*.patch
patches/ext4-mm-mballoc-core.patch
patches/git-kvm.patch
patches/git-perfmon.patch
patches/relay-reset-consumed.patch
patches/x86-andi-git-x86.patch
patches/x86-andi-smp-switch-optimize.patch
that's a sample of 1852 patches.
An appropriate warning would be "Usage of NR_CPUS is often wrong - should
you be using cpu_possible(), num_possible_cpus(), for_each_possible_cpu(),
etc?".
On Wed, Mar 05, 2008 at 02:12:29PM -0800, Andrew Morton wrote:
> On Mon, 03 Mar 2008 15:51:50 -0800
> "David J. Wilder" <[email protected]> wrote:
>
> > This patch allows relay channels to be reset i.e. unconsumed.
> > Basically allows a 'rewind' function for flight-recorder tracing.
> >
> > +void relay_reset_consumed(struct rchan *chan)
> > +{
> > + unsigned int i;
> > + struct rchan_buf *prev = NULL;
> > +
> > + if (!chan)
> > + return;
> > +
> > + for (i = 0; i < NR_CPUS; i++) {
>
> Use of NR_CPUS is usually wrong. In this case it seems you should be using
> for_each_possible_cpu().
>
> Also the existing usage of NR_CPUS in relay_subbufs_consumed() should be
> switched to using cpu_possible().
>
>
> New usage of NR_CPUS might be checkpatch-worthy, actually:
>
> akpm:/usr/src/25> grep -l '^+.*NR_CPUS' patches/*.patch
> patches/ext4-mm-mballoc-core.patch
> patches/git-kvm.patch
> patches/git-perfmon.patch
> patches/relay-reset-consumed.patch
> patches/x86-andi-git-x86.patch
> patches/x86-andi-smp-switch-optimize.patch
>
> that's a sample of 1852 patches.
>
> An appropriate warning would be "Usage of NR_CPUS is often wrong - should
> you be using cpu_possible(), num_possible_cpus(), for_each_possible_cpu(),
> etc?".
That seems reasonable. If we special case the definitions of NR_CPUS
and the use of them to define arrays of things both of which appear
reasonable,
#define NR_CPUS 10
int semid[NR_CPUS];
then it seems sensible to report this one:
WARNING: usage of NR_CPUS is often wrong - consider using cpu_possible(), num_possible_cpus(), for_each_possible_cpu(), etc
#1: FILE: Z101.c:1:
+ for (i = 0; i < NR_CPUS; i++)
This will be in the next push to -mm.
-apw