From: "Zheng, Lv" <lv.zheng@intel.com>
To: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: "Wysocki, Rafael J" <rafael.j.wysocki@intel.com>,
"Brown, Len" <len.brown@intel.com>, Lv Zheng <zetalog@gmail.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"linux-acpi@vger.kernel.org" <linux-acpi@vger.kernel.org>,
"Williams, Dan J" <dan.j.williams@intel.com>
Subject: RE: [PATCH v4 2/4] ACPICA: Tables: Add mechanism to allow to balance late stage acpi_get_table() independently
Date: Mon, 15 May 2017 06:32:48 +0000 [thread overview]
Message-ID: <1AE640813FDE7649BE1B193DEA596E886CEA3A13@SHSMSX101.ccr.corp.intel.com> (raw)
In-Reply-To: <1660120.KNfsmTX9Wg@aspire.rjw.lan>
Hi, Rafael
> From: Rafael J. Wysocki [mailto:rjw@rjwysocki.net]
> Subject: Re: [PATCH v4 2/4] ACPICA: Tables: Add mechanism to allow to balance late stage
> acpi_get_table() independently
>
> On Tuesday, May 09, 2017 01:57:41 PM Lv Zheng wrote:
> > For all frequent late stage acpi_get_table() clone invocations, we should
> > only change them altogether, otherwise, excessive acpi_put_table() could
> > unexpectedly unmap the table used by the other users. Thus the current plan
> > is to change all acpi_get_table() clones together or to change none of
> > them. However in practical, this is not convenient as this can prevent
> > kernel developers' efforts of improving the late stage code quality before
> > waiting for the ACPICA upstream to improve first.
> >
> > This patch adds a validation count threashold, when it is reached, the
> > validation count can no longer be incremented/decremented to invalidate the
> > table descriptor (means preventing table unmappings) so that acpi_put_table()
> > balance changes can be done independently to each others. Lv Zheng.
> >
> > Cc: Dan Williams <dan.j.williams@intel.com>
> > Signed-off-by: Lv Zheng <lv.zheng@intel.com>
> > ---
> > drivers/acpi/acpica/tbutils.c | 24 +++++++++++++++---------
> > include/acpi/actbl.h | 9 +++++++++
> > 2 files changed, 24 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/acpi/acpica/tbutils.c b/drivers/acpi/acpica/tbutils.c
> > index 7abe665..04beafc 100644
> > --- a/drivers/acpi/acpica/tbutils.c
> > +++ b/drivers/acpi/acpica/tbutils.c
> > @@ -416,9 +416,13 @@ acpi_tb_get_table(struct acpi_table_desc *table_desc,
> > }
> > }
> >
> > - table_desc->validation_count++;
> > - if (table_desc->validation_count == 0) {
> > - table_desc->validation_count--;
> > + if (table_desc->validation_count < ACPI_MAX_TABLE_VALIDATIONS) {
> > + table_desc->validation_count++;
> > + if (table_desc->validation_count >= ACPI_MAX_TABLE_VALIDATIONS) {
> > + ACPI_WARNING((AE_INFO,
> > + "Table %p, Validation count overflows\n",
> > + table_desc));
> > + }
> > }
> >
> > *out_table = table_desc->pointer;
> > @@ -445,13 +449,15 @@ void acpi_tb_put_table(struct acpi_table_desc *table_desc)
> >
> > ACPI_FUNCTION_TRACE(acpi_tb_put_table);
> >
> > - if (table_desc->validation_count == 0) {
> > - ACPI_WARNING((AE_INFO,
> > - "Table %p, Validation count is zero before decrement\n",
> > - table_desc));
> > - return_VOID;
> > + if (table_desc->validation_count < ACPI_MAX_TABLE_VALIDATIONS) {
> > + table_desc->validation_count--;
> > + if (table_desc->validation_count >= ACPI_MAX_TABLE_VALIDATIONS) {
>
> Is this going to ever trigger?
>
> We've already verified that validation_count is not 0 and that it is less than
> ACPI_MAX_TABLE_VALIDATIONS and we have decremented it, so how can it be
> greater than or equal to ACPI_MAX_TABLE_VALIDATIONS here?
This is just a no-op change equivalent to
"
if (validation_count == 0) { warn and return }
decrement
"
It expands "decrement" to "validation_count == 0" case so that it can implement warn_once for the warning message.
See:
A. validation_count == 0:
A.1. "if (validation_count < ACPI_MAX_TABLE_VALIDATIONS)" matches, and
After decrementing validation_count, it will be "0xFFFF";
Then "if (validation_count >= ACPI_MAX_TABLE_VALIDATIONS)" matches as "validation_count == 0xFFFF(ACPI_MAX_TABLE_VALIDATIONS)" now;
The warning message is printed;
A.2. "if (validation_count == ACPI_MAX_TABLE_VALIDATIONS)" doesn't match as "validation_count == 0xFFFF(ACPI_MAX_TABLE_VALIDATIONS)" now
the rest of this function will be skipped just like return_VOID.
B. validation_count == 0xFFFF:
A.1. Both acpi_tb_get_table() and acpi_tb_put_table() won't be able to change validation_count as
validation_count increment/decrement code fragments are only executed "if (validation_count < ACPI_MAX_TABLE_VALIDATIONS)"
Thus validation_count is kept as 0xFFFF (in this case, overflowed/underflowed values are same).
A.2. "if (validation_count == ACPI_MAX_TABLE_VALIDATIONS)" doesn't match as "validation_count == 0xFFFF" now
the rest of this function will be skipped just like return_VOID.
C. otherwise, validation_count will be decremented like old code
Benefits of using the new algorithm are:
1. ACPI_MAX_TABLE_VALIDATIONS can be something other than 0xFFFF.
It can be anything now, and the expected behavior can always be ensured.
IOW, the new algorithm actually supports cases where overflowed/underflowed values are not same.
You can check this by defining ACPI_MAX_TABLE_VALIDATIONS to 8.
And you'll see that the 2 functions are still working.
1.1. If something is broken in acpi_tb_get_table(), validation_count will be kept as 0x0008.
1.2. If something is broken in acpi_tb_put_table(), validation_count will be kept as 0xFFFF.
Both 0x0008 and 0xFFFF cannot make
"if (validation_count < ACPI_MAX_TABLE_VALIDATIONS)" and
"if (validation_count == 0)" to return true, and
Thus validation_count is kept unchanged after overflow/underflow.
2. The key benefit of this change is to make the old warning in acpi_tb_put_table() as warn_once.
For example:
acpi_get_table();
for (i = 0; i < 100; i++)
acpi_put_table()
Using the old algorithm, the acpi_tb_put_table() warning message will be seen 99 times.
Using the new algorithm, the acpi_tb_put_table() warning message will be seen only once.
3. logics in acpi_tb_put_table() will be exactly the reversal of the logics in acpi_tb_get_table().
It'll be easier to maintain both of them with the new overflow/underflow algorithm.
Hope you'll like such a change.
Thanks and best regards
Lv
>
> > + ACPI_WARNING((AE_INFO,
> > + "Table %p, Validation count underflows\n",
> > + table_desc));
> > + return_VOID;
> > + }
> > }
> > - table_desc->validation_count--;
> >
> > if (table_desc->validation_count == 0) {
> >
>
> Thanks,
> Rafael
next prev parent reply other threads:[~2017-05-15 6:32 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-04-27 8:22 [PATCH v2 1/2] ACPICA: Tables: Fix regression introduced by a too early mechanism enabling Lv Zheng
2017-04-27 8:22 ` [PATCH v2 2/2] ACPI: Fix memory mapping leaks in current sysfs dumpable ACPI tables support Lv Zheng
2017-04-27 22:32 ` Rafael J. Wysocki
2017-04-27 22:30 ` [PATCH v2 1/2] ACPICA: Tables: Fix regression introduced by a too early mechanism enabling Rafael J. Wysocki
2017-04-28 1:24 ` Zheng, Lv
2017-04-28 3:57 ` Zheng, Lv
2017-04-28 5:27 ` [PATCH v2 1/4] " Lv Zheng
2017-04-28 5:28 ` [PATCH v3 " Lv Zheng
2017-04-28 5:30 ` [PATCH v3 2/4] ACPICA: Tables: Add mechanism to allow to balance late stage acpi_get_table() independently Lv Zheng
2017-04-28 20:56 ` Rafael J. Wysocki
2017-05-04 7:18 ` Zheng, Lv
2017-05-04 15:45 ` Dan Williams
2017-05-05 0:53 ` Zheng, Lv
2017-05-05 20:43 ` Rafael J. Wysocki
2017-05-09 1:58 ` Zheng, Lv
2017-04-28 5:30 ` [PATCH v3 3/4] ACPI: sysfs: Fix acpi_get_table() leak Lv Zheng
2017-04-28 5:30 ` [PATCH v3 4/4] ACPI: Fix memory mapping leaks in current sysfs dumpable ACPI tables support Lv Zheng
2017-05-09 5:57 ` [PATCH v4 1/4] ACPICA: Tables: Fix regression introduced by a too early mechanism enabling Lv Zheng
2017-05-09 5:57 ` [PATCH v4 2/4] ACPICA: Tables: Add mechanism to allow to balance late stage acpi_get_table() independently Lv Zheng
2017-05-12 21:03 ` Rafael J. Wysocki
2017-05-12 21:41 ` Rafael J. Wysocki
2017-05-15 6:32 ` Zheng, Lv [this message]
2017-05-09 5:57 ` [PATCH v4 3/4] ACPI: sysfs: Fix acpi_get_table() leak Lv Zheng
2017-05-09 5:57 ` [PATCH v4 4/4] ACPI: Fix memory mapping leaks in current sysfs dumpable ACPI tables support Lv Zheng
2017-06-12 13:12 ` Rafael J. Wysocki
2017-06-07 4:54 ` [PATCH v5] ACPICA: Tables: Add mechanism to allow to balance late stage acpi_get_table() independently Lv Zheng
2017-06-07 6:41 ` Dan Williams
2017-06-07 21:14 ` Rafael J. Wysocki
2017-06-07 21:24 ` Dan Williams
2017-06-08 2:24 ` Zheng, Lv
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=1AE640813FDE7649BE1B193DEA596E886CEA3A13@SHSMSX101.ccr.corp.intel.com \
--to=lv.zheng@intel.com \
--cc=dan.j.williams@intel.com \
--cc=len.brown@intel.com \
--cc=linux-acpi@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=rafael.j.wysocki@intel.com \
--cc=rjw@rjwysocki.net \
--cc=zetalog@gmail.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).