From: Joel Fernandes <joelaf@google.com>
To: linux-kernel@vger.kernel.org
Cc: Steven Rostedt <rostedt@goodmis.org>,
Joel Fernandes <joelaf@google.com>,
Anton Vorontsov <anton@enomsg.org>,
Colin Cross <ccross@android.com>,
Kees Cook <keescook@chromium.org>,
Tony Luck <tony.luck@intel.com>
Subject: [PATCH 5/7] ramoops: Split ftrace buffer space into per-CPU zones
Date: Fri, 7 Oct 2016 22:28:32 -0700 [thread overview]
Message-ID: <1475904515-24970-6-git-send-email-joelaf@google.com> (raw)
In-Reply-To: <1475904515-24970-1-git-send-email-joelaf@google.com>
If FTRACE_PER_CPU flag is passed to ramoops pdata, split the space into
multiple zones depending on the number of CPUs.
This speeds up the performance of function tracing by about 280% in my tests as
we avoid the locking. The trade off being lesser space available per CPU. Let
the ramoops user decide which option they want based on pdata flag.
Signed-off-by: Joel Fernandes <joelaf@google.com>
---
fs/pstore/ram.c | 70 +++++++++++++++++++++++++++++++++++-----------
include/linux/pstore_ram.h | 3 ++
2 files changed, 56 insertions(+), 17 deletions(-)
diff --git a/fs/pstore/ram.c b/fs/pstore/ram.c
index a796f49..2e29d6b 100644
--- a/fs/pstore/ram.c
+++ b/fs/pstore/ram.c
@@ -87,7 +87,7 @@ MODULE_PARM_DESC(ramoops_ecc,
struct ramoops_context {
struct persistent_ram_zone **przs;
struct persistent_ram_zone *cprz;
- struct persistent_ram_zone *fprz;
+ struct persistent_ram_zone **fprzs;
struct persistent_ram_zone *mprz;
phys_addr_t phys_addr;
unsigned long size;
@@ -97,6 +97,7 @@ struct ramoops_context {
size_t ftrace_size;
size_t pmsg_size;
int dump_oops;
+ int flags;
struct persistent_ram_ecc_info ecc_info;
unsigned int max_dump_cnt;
unsigned int dump_write_cnt;
@@ -219,9 +220,17 @@ static ssize_t ramoops_pstore_read(u64 *id, enum pstore_type_id *type,
if (!prz_ok(prz))
prz = ramoops_get_next_prz(&cxt->cprz, &cxt->console_read_cnt,
1, id, type, PSTORE_TYPE_CONSOLE, 0);
- if (!prz_ok(prz))
- prz = ramoops_get_next_prz(&cxt->fprz, &cxt->ftrace_read_cnt,
- 1, id, type, PSTORE_TYPE_FTRACE, 0);
+ if (!prz_ok(prz)) {
+ int max = (cxt->flags & FTRACE_PER_CPU) ? nr_cpu_ids : 1;
+ while (cxt->ftrace_read_cnt < max && !prz) {
+ prz = ramoops_get_next_prz(cxt->fprzs,
+ &cxt->ftrace_read_cnt, max, id, type,
+ PSTORE_TYPE_FTRACE, 0);
+ if (!prz_ok(prz))
+ continue;
+ }
+ }
+
if (!prz_ok(prz))
prz = ramoops_get_next_prz(&cxt->mprz, &cxt->pmsg_read_cnt,
1, id, type, PSTORE_TYPE_PMSG, 0);
@@ -283,9 +292,23 @@ static int notrace ramoops_pstore_write_buf(enum pstore_type_id type,
persistent_ram_write(cxt->cprz, buf, size, 1);
return 0;
} else if (type == PSTORE_TYPE_FTRACE) {
- if (!cxt->fprz)
+ int zonenum, lock;
+
+ if (!cxt->fprzs)
return -ENOMEM;
- persistent_ram_write(cxt->fprz, buf, size, 1);
+ /*
+ * If per-cpu buffers, don't lock. Otherwise there's only
+ * 1 zone for ftrace (zone 0) and all CPUs share it, so lock.
+ */
+ if (cxt->flags & FTRACE_PER_CPU) {
+ zonenum = smp_processor_id();
+ lock = 0;
+ } else {
+ zonenum = 0;
+ lock = 1;
+ }
+
+ persistent_ram_write(cxt->fprzs[zonenum], buf, size, lock);
return 0;
}
@@ -349,6 +372,7 @@ static int ramoops_pstore_erase(enum pstore_type_id type, u64 id, int count,
{
struct ramoops_context *cxt = psi->data;
struct persistent_ram_zone *prz;
+ int max;
switch (type) {
case PSTORE_TYPE_DMESG:
@@ -360,7 +384,10 @@ static int ramoops_pstore_erase(enum pstore_type_id type, u64 id, int count,
prz = cxt->cprz;
break;
case PSTORE_TYPE_FTRACE:
- prz = cxt->fprz;
+ max = (cxt->flags & FTRACE_PER_CPU) ? nr_cpu_ids : 1;
+ if (id >= max)
+ return -EINVAL;
+ prz = cxt->fprzs[id];
break;
case PSTORE_TYPE_PMSG:
prz = cxt->mprz;
@@ -391,14 +418,21 @@ static void ramoops_free_przs(struct ramoops_context *cxt)
{
int i;
- if (!cxt->przs)
- return;
+ /* Free dump PRZs */
+ if (cxt->przs) {
+ for (i = 0; i < cxt->max_dump_cnt; i++)
+ persistent_ram_free(cxt->przs[i]);
- for (i = 0; i < cxt->max_dump_cnt; i++)
- persistent_ram_free(cxt->przs[i]);
+ kfree(cxt->przs);
+ cxt->max_dump_cnt = 0;
+ }
- kfree(cxt->przs);
- cxt->max_dump_cnt = 0;
+ /* Free ftrace PRZs */
+ if (cxt->fprzs) {
+ for (i = 0; i < nr_cpu_ids; i++)
+ persistent_ram_free(cxt->przs[i]);
+ kfree(cxt->fprzs);
+ }
}
static int ramoops_init_przs(struct device *dev, struct ramoops_context *cxt,
@@ -615,6 +649,7 @@ static int ramoops_probe(struct platform_device *pdev)
cxt->ftrace_size = pdata->ftrace_size;
cxt->pmsg_size = pdata->pmsg_size;
cxt->dump_oops = pdata->dump_oops;
+ cxt->flags = pdata->flags;
cxt->ecc_info = pdata->ecc_info;
paddr = cxt->phys_addr;
@@ -630,8 +665,9 @@ static int ramoops_probe(struct platform_device *pdev)
if (err)
goto fail_init_cprz;
- err = ramoops_init_prz(dev, cxt, &cxt->fprz, &paddr, cxt->ftrace_size,
- LINUX_VERSION_CODE);
+ err = ramoops_init_przs(dev, cxt, &cxt->fprzs, &paddr, cxt->ftrace_size,
+ (cxt->flags & FTRACE_PER_CPU) ? nr_cpu_ids : 1,
+ LINUX_VERSION_CODE);
if (err)
goto fail_init_fprz;
@@ -695,7 +731,6 @@ fail_clear:
cxt->pstore.bufsize = 0;
persistent_ram_free(cxt->mprz);
fail_init_mprz:
- persistent_ram_free(cxt->fprz);
fail_init_fprz:
persistent_ram_free(cxt->cprz);
fail_init_cprz:
@@ -714,7 +749,6 @@ static int ramoops_remove(struct platform_device *pdev)
cxt->pstore.bufsize = 0;
persistent_ram_free(cxt->mprz);
- persistent_ram_free(cxt->fprz);
persistent_ram_free(cxt->cprz);
ramoops_free_przs(cxt);
@@ -756,6 +790,8 @@ static void ramoops_register_dummy(void)
dummy_data->ftrace_size = ramoops_ftrace_size;
dummy_data->pmsg_size = ramoops_pmsg_size;
dummy_data->dump_oops = dump_oops;
+ dummy_data->flags = FTRACE_PER_CPU;
+
/*
* For backwards compatibility ramoops.ecc=1 means 16 bytes ECC
* (using 1 byte for ECC isn't much of use anyway).
diff --git a/include/linux/pstore_ram.h b/include/linux/pstore_ram.h
index 782af68..a30573b 100644
--- a/include/linux/pstore_ram.h
+++ b/include/linux/pstore_ram.h
@@ -78,6 +78,8 @@ ssize_t persistent_ram_ecc_string(struct persistent_ram_zone *prz,
* @mem_address physical memory address to contain ramoops
*/
+#define FTRACE_PER_CPU BIT(0)
+
struct ramoops_platform_data {
unsigned long mem_size;
phys_addr_t mem_address;
@@ -87,6 +89,7 @@ struct ramoops_platform_data {
unsigned long ftrace_size;
unsigned long pmsg_size;
int dump_oops;
+ int flags;
struct persistent_ram_ecc_info ecc_info;
};
--
2.7.4
next prev parent reply other threads:[~2016-10-08 5:30 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-10-08 5:28 [RFC 0/7] pstore: Improve performance of ftrace backend with ramoops Joel Fernandes
2016-10-08 5:28 ` [PATCH 1/7] pstore: Make spinlock per zone instead of global Joel Fernandes
2016-10-08 5:44 ` Joel Fernandes
2016-10-10 23:44 ` Kees Cook
2016-10-08 5:28 ` [PATCH 2/7] pstore: locking: dont lock unless caller asks to Joel Fernandes
2016-10-10 23:48 ` Kees Cook
2016-10-11 14:41 ` Joel Fernandes
2016-10-08 5:28 ` [PATCH 3/7] pstore: Remove case of PSTORE_TYPE_PMSG write using deprecated function Joel Fernandes
2016-10-10 23:52 ` Kees Cook
2016-10-11 14:46 ` Joel Fernandes
2016-10-08 5:28 ` [PATCH 4/7] pstore: Make ramoops_init_przs generic for other prz arrays Joel Fernandes
2016-10-10 23:55 ` Kees Cook
2016-10-08 5:28 ` Joel Fernandes [this message]
2016-10-09 17:15 ` [PATCH 5/7] ramoops: Split ftrace buffer space into per-CPU zones Joel Fernandes
2016-10-10 23:59 ` Kees Cook
2016-10-11 0:00 ` Kees Cook
2016-10-16 17:40 ` Joel Fernandes
2016-10-18 20:37 ` Kees Cook
2016-10-08 5:28 ` [PATCH 6/7] pstore: Add support to store timestamp counter in ftrace records Joel Fernandes
2016-10-08 5:28 ` [PATCH 7/7] pstore: Merge per-CPU ftrace zones into one zone for output Joel Fernandes
2016-10-11 9:57 ` [RFC 0/7] pstore: Improve performance of ftrace backend with ramoops Steven Rostedt
2016-10-20 7:17 [PATCH " Joel Fernandes
2016-10-20 7:17 ` [PATCH 5/7] ramoops: Split ftrace buffer space into per-CPU zones Joel Fernandes
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1475904515-24970-6-git-send-email-joelaf@google.com \
--to=joelaf@google.com \
--cc=anton@enomsg.org \
--cc=ccross@android.com \
--cc=keescook@chromium.org \
--cc=linux-kernel@vger.kernel.org \
--cc=rostedt@goodmis.org \
--cc=tony.luck@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).