Date:   Mon, 13 Feb 2017 14:32:53 +0100
From:   Peter Zijlstra <peterz@...radead.org>
To:     Kefeng Wang <wangkefeng.wang@...wei.com>
Cc:     Ingo Molnar <mingo@...nel.org>, linux-kernel@...r.kernel.org,
        guohanjun@...wei.com, Prarit Bhargava <prarit@...hat.com>,
        Tejun Heo <tj@...nel.org>
Subject: Re: [PATCH] sched/isolcpus: Show isolated cpu map

On Mon, Feb 13, 2017 at 09:07:02PM +0800, Kefeng Wang wrote:
> Hi Peter
> 
> +Tejun
> 
> On 2017/2/13 20:06, Peter Zijlstra wrote:
> > On Mon, Feb 13, 2017 at 05:43:59PM +0800, Kefeng Wang wrote:
> >> The commit a6e4491c682a ("sched/isolcpus: Output warning when the
> >> 'isolcpus=' kernel parameter is invalid") adds an error message
> >> when specified cpu bigger than nr_cpu_ids, but nr_cpumask_bits in
> >> cpulist_parse() could be nr_cpu_ids or NR_CPUS.
> >>
> >> eg, NR_CPUS=64, nr_cpu_ids=8 in ARM64, cpulist_parse() won't return
> >> -ERANGE if isolcpus=1-10;
> >>
> > 
> > But why does cpulist_parse() use nr_cpumask_bits, that seems to be the
> > problem, so why not look there?
> > 
> > 
> 
> Paste the Tejun's patch,
> 
> commit 4d59b6ccf000862beed6fc0765d3209f98a8d8a2
> Author: Tejun Heo <tj@...nel.org>
> Date:   Wed Feb 8 14:30:56 2017 -0800
> 
>     cpumask: use nr_cpumask_bits for parsing functions
> 
>     Commit 513e3d2d11c9 ("cpumask: always use nr_cpu_ids in formatting and
>     parsing functions") converted both cpumask printing and parsing
>     functions to use nr_cpu_ids instead of nr_cpumask_bits.  While this was
>     okay for the printing functions as it just picked one of the two output
>     formats that we were alternating between depending on a kernel config,
>     doing the same for parsing wasn't okay.
> 
>     nr_cpumask_bits can be either nr_cpu_ids or NR_CPUS.  We can always use
>     nr_cpu_ids but that is a variable while NR_CPUS is a constant, so it can
>     be more efficient to use NR_CPUS when we can get away with it.
>     Converting the printing functions to nr_cpu_ids makes sense because it
>     affects how the masks get presented to userspace and doesn't break
>     anything; however, using nr_cpu_ids for parsing functions can
>     incorrectly leave the higher bits uninitialized while reading in these
>     masks from userland.  As all testing and comparison functions use
>     nr_cpumask_bits which can be larger than nr_cpu_ids, the parsed cpumasks
>     can erroneously yield false negative results.
> 
>     This made the taskstats interface incorrectly return -EINVAL even when
>     the inputs were correct.
> 
>     Fix it by restoring the parse functions to use nr_cpumask_bits instead
>     of nr_cpu_ids.

OK, so its wrong both ways.

Problem seems to be that cpumask is internally inconsistent with the
number of bits because a small constant NR_CPUS is more efficient for
things like cpumask_subset().

If everything were consistent and used nr_cpu_ids it would all be fine,
but using a mixture is giving pain.

Does something like the below work? It parses up to nr_cpu_ids and then
and's with cpu_possible_mask (which has all bits set). In case
nr_cpumask_bits is larger than nr_cpu_ids this should result in clearing
the top bits (and therefore not leave them uninitialized). And using
nr_cpu_ids for parsing now makes the range check work again.

Since parsing in general is a really slow thing anyway, the extra
cpumask operation doesn't matter.

---
 include/linux/cpumask.h | 32 +++++++++++++++++++++++++++-----
 1 file changed, 27 insertions(+), 5 deletions(-)

diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
index 96f1e88b767c..6cf8945b999d 100644
--- a/include/linux/cpumask.h
+++ b/include/linux/cpumask.h
@@ -560,7 +560,13 @@ static inline void cpumask_copy(struct cpumask *dstp,
 static inline int cpumask_parse_user(const char __user *buf, int len,
 				     struct cpumask *dstp)
 {
-	return bitmap_parse_user(buf, len, cpumask_bits(dstp), nr_cpumask_bits);
+	int ret;
+
+	ret = bitmap_parse_user(buf, len, cpumask_bits(dstp), nr_cpu_ids);
+	if (!ret)
+		cpumask_and(dstp, dstp, cpu_possible_mask);
+
+	return ret;
 }
 
 /**
@@ -574,8 +580,13 @@ static inline int cpumask_parse_user(const char __user *buf, int len,
 static inline int cpumask_parselist_user(const char __user *buf, int len,
 				     struct cpumask *dstp)
 {
-	return bitmap_parselist_user(buf, len, cpumask_bits(dstp),
-				     nr_cpumask_bits);
+	int ret;
+
+	ret = bitmap_parselist_user(buf, len, cpumask_bits(dstp), nr_cpu_ids);
+	if (!ret)
+		cpumask_and(dstp, dstp, cpu_possible_mask);
+
+	return ret;
 }
 
 /**
@@ -589,8 +600,13 @@ static inline int cpumask_parse(const char *buf, struct cpumask *dstp)
 {
 	char *nl = strchr(buf, '\n');
 	unsigned int len = nl ? (unsigned int)(nl - buf) : strlen(buf);
+	int ret;
 
-	return bitmap_parse(buf, len, cpumask_bits(dstp), nr_cpumask_bits);
+	ret = bitmap_parse(buf, len, cpumask_bits(dstp), nr_cpu_ids);
+	if (!ret)
+		cpumask_and(dstp, dstp, cpu_possible_mask);
+
+	return ret;
 }
 
 /**
@@ -602,7 +618,13 @@ static inline int cpumask_parse(const char *buf, struct cpumask *dstp)
  */
 static inline int cpulist_parse(const char *buf, struct cpumask *dstp)
 {
-	return bitmap_parselist(buf, cpumask_bits(dstp), nr_cpumask_bits);
+	int ret;
+
+	ret = bitmap_parselist(buf, cpumask_bits(dstp), nr_cpu_ids);
+	if (!ret)
+		cpumask_and(dstp, dstp, cpu_possible_mask);
+
+	return ret;
 }
 
 /**

Powered by blists - more mailing lists