Skip to contentSkip to conversation

Fix incorrect int->float conversions caught by clang -Wimplicit-int-float-conversion

PATCH Superseded Fangrui Song 5 participants 11 messages Nov 16, 2019 – Nov 22, 2019

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix incorrect int->float conversions caught by clang -Wimplicit-int-float-conversion

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--
  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +2 -2
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
  #include "monitor/monitor.h"
  #include "net/announce.h"
  #include "qemu/queue.h"
+#include <math.h>
  
  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed throttling */
  
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
      }
  
      value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
  
      MigrateSetParameters p = {
          .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),
      };
  
      qmp_migrate_set_parameters(&p, errp);
util/cutils.c +2 -2
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
          goto out;
      }
      /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
       * through double (53 bits of precision).
       */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
          retval = -ERANGE;
          goto out;
      }
Signature · 1 line
2.24.0
#1 lore raw

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp)

Quoted 25 lines · Fangrui Song

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--
  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +1 -0
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
  #include "monitor/monitor.h"
  #include "net/announce.h"
  #include "qemu/queue.h"
+#include <math.h>
 
  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed throttling */
 
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
        if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) {
            error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                             "the range of 0 to %d seconds",
                             MAX_MIGRATE_DOWNTIME_SECONDS);
            return;

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2036

Quoted 1 line
migration/migration.c +0 -0
      }

@value is now in [0,2000].

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 2 lines
migration/migration.c +0 -0
      value *= 1000; /* Convert to milliseconds */

@value is in [0,2000000]

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 1 line
migration/migration.c +0 -1
-    value = MAX(0, MIN(INT64_MAX, value));

This does nothing.

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2042

Quoted 5 lines
migration/migration.c +1 -1
      MigrateSetParameters p = {
          .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),

This does nothing and is hard to read :)

Can we simply drop the offending line statement instead?

Quoted 10 lines
migration/migration.c +0 -0
      };
 
      qmp_migrate_set_parameters(&p, errp);
util/cutils.c +0 -0
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
          goto out;
      }

util/cutils.c static int do_strtosz(const char *nptr, const char **end, · line 248

Quoted 10 lines
util/cutils.c +2 -2
      /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
       * through double (53 bits of precision).
       */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
          retval = -ERANGE;
          goto out;
      }
        *result = val * mul;

I figure this one is correct and hard to read.

0xfffffffffffffc00 is not representable exactly as double. It's half-way between the representable values 0xfffffffffffff800 and 0x10000000000000000. Which one we get is implementation-defined. Bad.

nextafter(0x1p64, 0) is a clever way to write 0xfffffffffffff800, the largest uint64_t exactly representable as double.

With your patch, val * mul in [0,0xfffffffffffff800] will be accepted.

The first val * mul above this range is 0x1p64. Rejecting it is correct, because it overflows yint64_t.

#2 lore raw
Quoted 51 lines · Markus Armbruster

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp)

Quoted 30 lines · Fangrui Song

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--
  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +1 -0
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
  #include "monitor/monitor.h"
  #include "net/announce.h"
  #include "qemu/queue.h"
+#include <math.h>
 
  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed throttling */
 
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
        if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) {
            error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                             "the range of 0 to %d seconds",
                             MAX_MIGRATE_DOWNTIME_SECONDS);
            return;

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2036

Quoted 1 line
migration/migration.c +0 -0
      }

@value is now in [0,2000].

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 2 lines
migration/migration.c +0 -0
      value *= 1000; /* Convert to milliseconds */

@value is in [0,2000000]

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 1 line
migration/migration.c +0 -1
-    value = MAX(0, MIN(INT64_MAX, value));

This does nothing.

Quoted 1 line
Quoted 8 lines

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2042

Quoted 4 lines

MigrateSetParameters p = {
.has_downtime_limit = true,

migration/migration.c +1 -1
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),

This does nothing and is hard to read :)

Can we simply drop the offending line statement instead?

Agreed aboutdropping the whole bussines for migration.

Quoted 12 lines
Quoted 12 lines
      };
  
      qmp_migrate_set_parameters(&p, errp);
util/cutils.c +0 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
          goto out;
      }
      /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
Quoted 8 lines

util/cutils.c static int do_strtosz(const char *nptr, const char **end, · line 248

Quoted 8 lines
util/cutils.c +2 -1
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
       * through double (53 bits of precision).
       */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
          retval = -ERANGE;
          goto out;
      }

This comment was really bad (it says the same that the code). On the other hand, I can *kind of* understand what does 0xffff<more f's here>.

But I am at a complete loss about what value is:

nextafter(0x1p64, 0).

Can we put what value is that instead?

Thanks, Juan.

#3 lore raw
Quoted 89 lines · Juan Quintela
Quoted 51 lines · Markus Armbruster

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp)

Quoted 30 lines · Fangrui Song

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--
  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +1 -0
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
  #include "monitor/monitor.h"
  #include "net/announce.h"
  #include "qemu/queue.h"
+#include <math.h>
 
  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed throttling */
 
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
        if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) {
            error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                             "the range of 0 to %d seconds",
                             MAX_MIGRATE_DOWNTIME_SECONDS);
            return;

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2036

Quoted 1 line
migration/migration.c +0 -0
      }

@value is now in [0,2000].

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 2 lines
migration/migration.c +0 -0
      value *= 1000; /* Convert to milliseconds */

@value is in [0,2000000]

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2038

Quoted 1 line
migration/migration.c +0 -1
-    value = MAX(0, MIN(INT64_MAX, value));

This does nothing.

Quoted 1 line
Quoted 8 lines

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2042

Quoted 4 lines

MigrateSetParameters p = {
.has_downtime_limit = true,

migration/migration.c +1 -1
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),

This does nothing and is hard to read :)

Can we simply drop the offending line statement instead?

Agreed aboutdropping the whole bussines for migration.

Quoted 12 lines
Quoted 12 lines
      };

      qmp_migrate_set_parameters(&p, errp);
util/cutils.c +0 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
          goto out;
      }
      /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
Quoted 8 lines

util/cutils.c static int do_strtosz(const char *nptr, const char **end, · line 248

Quoted 8 lines
util/cutils.c +2 -1
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
       * through double (53 bits of precision).
       */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
          retval = -ERANGE;
          goto out;
      }

This comment was really bad (it says the same that the code). On the other hand, I can *kind of* understand what does 0xffff<more f's here>.

Quoted 5 lines

But I am at a complete loss about what value is:

nextafter(0x1p64, 0).

Can we put what value is that instead?

It is a C99 hexadecimal floating-point literal.
0x1p64 represents hex fraction 1.0 scaled by 2**64, that is 2**64.

We can write this as val * mul > 0xfffffffffffff800p0, but I feel that counting the number of f's is error-prone and is not fun.

(We cannot use val * mul >= 0x1p64.
If FLT_EVAL_METHOD == 2, the intermediate computation val * mul will be performed at long double precision, val * mul may not by representable by a double and will overflow as (double)0x1p64.)

#4 lore raw
Quoted 103 lines · Fangrui Song
Quoted 93 lines · Juan Quintela
Quoted 54 lines · Markus Armbruster
Quoted 31 lines · Fangrui Song

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--

  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +1 -0
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
   #include "monitor/monitor.h"
   #include "net/announce.h"
   #include "qemu/queue.h"
+#include <math.h>
 
   #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed

throttling */

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2036

Quoted 1 line
migration/migration.c +0 -0
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error
 **errp)

        if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) { error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                             "the range of 0 to %d seconds", MAX_MIGRATE_DOWNTIME_SECONDS); return;

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2037

Quoted 1 line
migration/migration.c +0 -0
       }

@value is now in [0,2000].

Quoted 2 lines

      value *= 1000; /* Convert to milliseconds */

@value is in [0,2000000]

Quoted 1 line

-    value = MAX(0, MIN(INT64_MAX, value));

This does nothing.

Quoted 1 line
Quoted 8 lines
Quoted 4 lines

      MigrateSetParameters p = {
          .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),

This does nothing and is hard to read :)

Can we simply drop the offending line statement instead?

Agreed aboutdropping the whole bussines for migration.

Quoted 13 lines
Quoted 13 lines

      };

      qmp_migrate_set_parameters(&p, errp);

util/cutils.c +0 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char
 **end,
           goto out;
       }
       /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
Quoted 8 lines
Quoted 8 lines
util/cutils.c +2 -1
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
        * through double (53 bits of precision).
        */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
           retval = -ERANGE;
           goto out;

      }

This comment was really bad (it says the same that the code). On the other hand, I can *kind of* understand what does 0xffff<more f's here>.

Quoted 5 lines

But I am at a complete loss about what value is:

nextafter(0x1p64, 0).

Can we put what value is that instead?

It is a C99 hexadecimal floating-point literal.
0x1p64 represents hex fraction 1.0 scaled by 2**64, that is 2**64.

Quoted 7 lines

We can write this as val * mul > 0xfffffffffffff800p0, but I feel that counting the number of f's is error-prone and is not fun.

(We cannot use val * mul >= 0x1p64.
If FLT_EVAL_METHOD == 2, the intermediate computation val * mul will be performed at long double precision, val * mul may not by representable by a double and will overflow as (double)0x1p64.)

I agree about not spelling out the f's, or the 0x800 at the end. That's something that the compiler can do for us, resolving this standard library function at compile-time.

We just need a better comment. Perhaps:

/*
* Values near UINT64_MAX overflow to 2**64 when converting
* to double precision. Compare against the maximum representable
* double precision value below 2**64, computed as "the next value
* after 2**64 (0x1p64) in the direction of 0".
*/

r~

#5 lore raw
Quoted 4 lines · Richard Henderson
Quoted 3 lines · Fangrui Song
Quoted 2 lines · Juan Quintela
Quoted 1 line · Markus Armbruster

Fangrui Song <[email protected]> writes:

[...]

Quoted 44 lines
Quoted 32 lines
Quoted 23 lines
Quoted 10 lines
Quoted 10 lines
util/cutils.c +0 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char
 **end,
           goto out;
       }
       /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
Quoted 8 lines
Quoted 8 lines
util/cutils.c +2 -1
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
        * through double (53 bits of precision).
        */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
           retval = -ERANGE;
           goto out;

      }

This comment was really bad (it says the same that the code). On the other hand, I can *kind of* understand what does 0xffff<more f's here>.

Quoted 5 lines

But I am at a complete loss about what value is:

nextafter(0x1p64, 0).

Can we put what value is that instead?

It is a C99 hexadecimal floating-point literal.
0x1p64 represents hex fraction 1.0 scaled by 2**64, that is 2**64.

Quoted 7 lines

We can write this as val * mul > 0xfffffffffffff800p0, but I feel that counting the number of f's is error-prone and is not fun.

(We cannot use val * mul >= 0x1p64.
If FLT_EVAL_METHOD == 2, the intermediate computation val * mul will be performed at long double precision, val * mul may not by representable by a double and will overflow as (double)0x1p64.)

I agree about not spelling out the f's, or the 0x800 at the end. That's something that the compiler can do for us, resolving this standard library function at compile-time.

Quoted 8 lines

We just need a better comment. Perhaps:

/*
* Values near UINT64_MAX overflow to 2**64 when converting
* to double precision. Compare against the maximum representable
* double precision value below 2**64, computed as "the next value
* after 2**64 (0x1p64) in the direction of 0".
*/

Yes, please.

#6 lore raw
Quoted 53 lines · Markus Armbruster
Quoted 4 lines · Richard Henderson
Quoted 3 lines · Fangrui Song
Quoted 2 lines · Juan Quintela
Quoted 1 line · Markus Armbruster

Fangrui Song <[email protected]> writes:

[...]

Quoted 46 lines
Quoted 32 lines
Quoted 23 lines
Quoted 10 lines
Quoted 10 lines
util/cutils.c +0 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char
 **end,
           goto out;
       }
       /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
Quoted 8 lines
Quoted 8 lines
util/cutils.c +2 -1
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
        * through double (53 bits of precision).
        */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
           retval = -ERANGE;
           goto out;

      }

This comment was really bad (it says the same that the code). On the other hand, I can *kind of* understand what does 0xffff<more f's here>.

Quoted 5 lines

But I am at a complete loss about what value is:

nextafter(0x1p64, 0).

Can we put what value is that instead?

It is a C99 hexadecimal floating-point literal.
0x1p64 represents hex fraction 1.0 scaled by 2**64, that is 2**64.

Quoted 7 lines

We can write this as val * mul > 0xfffffffffffff800p0, but I feel that counting the number of f's is error-prone and is not fun.

(We cannot use val * mul >= 0x1p64.
If FLT_EVAL_METHOD == 2, the intermediate computation val * mul will be performed at long double precision, val * mul may not by representable by a double and will overflow as (double)0x1p64.)

I agree about not spelling out the f's, or the 0x800 at the end. That's something that the compiler can do for us, resolving this standard library function at compile-time.

We just need a better comment. Perhaps:

Quoted 8 lines
Quoted 6 lines

/*
* Values near UINT64_MAX overflow to 2**64 when converting
* to double precision. Compare against the maximum representable
* double precision value below 2**64, computed as "the next value
* after 2**64 (0x1p64) in the direction of 0".
*/

Yes, please.

Thanks for the suggestion. Attached a new patch.

From 13312e91e5565a6bed8c394d5711603c7a8f8a3c Mon Sep 17 00:00:00 2001
From: Fangrui Song <[email protected]>
Date: Fri, 15 Nov 2019 16:27:47 -0800
Subject: [PATCH] Fix incorrect integer->float conversion caught by clang -Wimplicit-int-float-conversion
To: [email protected]

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
 migration/migration.c | 3 +--
 util/cutils.c         | 8 +++++---
 2 files changed, 6 insertions(+), 5 deletions(-)
migration/migration.c +1 -2
index 354ad072fa..09b150663f 100644
@@ -2035,11 +2035,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
     }
 
     value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
 
     MigrateSetParameters p = {
         .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)value,
     };
 
     qmp_migrate_set_parameters(&p, errp);
util/cutils.c +5 -3
index fd591cadf0..77acadc70a 100644
@@ -239,10 +239,12 @@ static int do_strtosz(const char *nptr, const char **end,
         goto out;
     }
     /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
-     * through double (53 bits of precision).
+     * Values near UINT64_MAX overflow to 2**64 when converting to double
+     * precision.  Compare against the maximum representable double precision
+     * value below 2**64, computed as "the next value after 2**64 (0x1p64) in
+     * the direction of 0".
      */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
         retval = -ERANGE;
         goto out;
     }
Signature · 1 line
2.24.0
#7 lore raw
In reply to Fangrui Song
Quoted 49 lines
Quoted 26 lines · Fangrui Song

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
  migration/migration.c | 4 ++--
  util/cutils.c         | 4 ++--
  2 files changed, 4 insertions(+), 4 deletions(-)
migration/migration.c +1 -0
index 354ad072fa..ac3ea2934a 100644
@@ -53,6 +53,7 @@
  #include "monitor/monitor.h"
  #include "net/announce.h"
  #include "qemu/queue.h"
+#include <math.h>
 
  #define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed 

throttling */

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2036

Quoted 1 line
migration/migration.c +0 -0
@@ -2035,11 +2036,10 @@ void qmp_migrate_set_downtime(double value, Error 
 **errp)
        if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) {
            error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                             "the range of 0 to %d seconds",
                             MAX_MIGRATE_DOWNTIME_SECONDS);
            return;

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2037

Quoted 1 line
migration/migration.c +0 -0
      }

@value is now in [0,2000].

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2038

Quoted 2 lines
migration/migration.c +0 -0
      value *= 1000; /* Convert to milliseconds */

@value is in [0,2000000]

migration/migration.c void qmp_migrate_set_downtime(double value, Error · line 2038

Quoted 1 line
migration/migration.c +0 -1
-    value = MAX(0, MIN(INT64_MAX, value));

This does nothing.

Quoted 1 line
Quoted 8 lines

migration/migration.c

Quoted 4 lines

MigrateSetParameters p = {
.has_downtime_limit = true,

migration/migration.c +1 -1
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)fmin(value, nextafter(0x1p63, 0)),

This does nothing and is hard to read :)

Can we simply drop the offending line statement instead?

Fixed in the new patch.

Quoted 29 lines
Quoted 20 lines
      };
  
      qmp_migrate_set_parameters(&p, errp);
util/cutils.c +2 -2
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char 
 **end,
          goto out;
      }
      /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
       * through double (53 bits of precision).
       */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
          retval = -ERANGE;
          goto out;
Quoted 1 line
      }
        *result = val * mul;

I figure this one is correct and hard to read.

0xfffffffffffffc00 is not representable exactly as double. It's half-way between the representable values 0xfffffffffffff800 and 0x10000000000000000. Which one we get is implementation-defined. Bad.

Quoted 7 lines

nextafter(0x1p64, 0) is a clever way to write 0xfffffffffffff800, the largest uint64_t exactly representable as double.

With your patch, val * mul in [0,0xfffffffffffff800] will be accepted.

The first val * mul above this range is 0x1p64. Rejecting it is correct, because it overflows yint64_t.

I am not subscribed, so apologize that this email may be off the thread.

(The binutils mailing list allows a user to download the raw email so I can still reply to a specific email, but this list does not provide such feature.)

From 5f1c5a42794ddcbabb63d9af920d9f437ea90a9f Mon Sep 17 00:00:00 2001
From: Fangrui Song <[email protected]>
Date: Fri, 15 Nov 2019 16:27:47 -0800
Subject: [PATCH] Fix incorrect integer->float conversions caught by clang
-Wimplicit-int-float-conversion
To: [email protected]

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
 migration/migration.c | 3 +--
 util/cutils.c         | 4 ++--
 2 files changed, 3 insertions(+), 4 deletions(-)
migration/migration.c +1 -2
index 354ad072fa..09b150663f 100644
@@ -2035,11 +2035,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
     }
 
     value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
 
     MigrateSetParameters p = {
         .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)value,
     };
 
     qmp_migrate_set_parameters(&p, errp);
util/cutils.c +2 -2
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
         goto out;
     }
     /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
      * through double (53 bits of precision).
      */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
         retval = -ERANGE;
         goto out;
     }
Signature · 1 line
2.24.0
#8 lore raw
Quoted 6 lines · Fangrui Song
Quoted 2 lines

Can we simply drop the offending line statement instead?

Fixed in the new patch.

Quoted 8 lines
Quoted 2 lines

The first val * mul above this range is 0x1p64.  Rejecting it is correct, because it overflows yint64_t.

I am not subscribed, so apologize that this email may be off the thread.

(The binutils mailing list allows a user to download the raw email so I can still reply to a specific email, but this list does not provide such feature.)

Actually, it's better to post a v2 patch as a new top-level thread, rather than buried as an attachment to a reply to v1, because our CI tooling doesn't see through the attachment (nor was it easy for me to reply to the v2 patch - I had to open the attachment to paste its text inline below...).

More patch submission hints at https://wiki.qemu.org/Contribute/SubmitAPatch

Quoted 12 lines
Quoted 1 line

From 5f1c5a42794ddcbabb63d9af920d9f437ea90a9f Mon Sep 17 00:00:00 2001

From: Fangrui Song <[email protected]>
Date: Fri, 15 Nov 2019 16:27:47 -0800
Subject: [PATCH] Fix incorrect integer->float conversions caught by clang
-Wimplicit-int-float-conversion
To: [email protected]

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

+++ b/migration/migration.c

migration/migration.c void qmp_migrate_set_downtime(double value, Error **errp) · line 2042

Quoted 10 lines
migration/migration.c +1 -2
@@ -2035,11 +2035,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
     }
 
     value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
 
     MigrateSetParameters p = {
         .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)value,
     };

The explicit cast looks odd without a comment (generally, we try to avoid casts, so a comment such as /* explicit cast to silence compiler */ can be useful)

util/cutils.c static int do_strtosz(const char *nptr, const char **end, · line 243

Quoted 13 lines
migration/migration.c +0 -0
     qmp_migrate_set_parameters(&p, errp);
util/cutils.c +1 -1
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
         goto out;
     }
     /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
      * through double (53 bits of precision).

I thought we agreed on more text than just this (in particular, that the nextafter() call represents 2^64 rounded towards zero).

Quoted 8 lines
util/cutils.c +1 -1
      */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
         retval = -ERANGE;
         goto out;
     }
Signature · 1 line
2.24.0
Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
#9 lore raw

PATCH v2 1/1 Fix incorrect integer->float conversions caught by clang -Wimplicit-int-float-conversion

Quoted 13 lines · Eric Blake
Quoted 11 lines · Fangrui Song
Quoted 2 lines

Can we simply drop the offending line statement instead?

Fixed in the new patch.

Quoted 2 lines

The first val * mul above this range is 0x1p64.  Rejecting it is correct, because it overflows yint64_t.

I am not subscribed, so apologize that this email may be off the thread.

Quoted 7 lines
Quoted 3 lines

(The binutils mailing list allows a user to download the raw email so I can still reply to a specific email, but this list does not provide such feature.)

Actually, it's better to post a v2 patch as a new top-level thread, rather than buried as an attachment to a reply to v1, because our CI tooling doesn't see through the attachment (nor was it easy for me to reply to the v2 patch - I had to open the attachment to paste its text inline below...).

More patch submission hints at https://wiki.qemu.org/Contribute/SubmitAPatch

Retitled to [PATCH v2]

Quoted 10 lines
Quoted 9 lines
Quoted 1 line

From 5f1c5a42794ddcbabb63d9af920d9f437ea90a9f Mon Sep 17 00:00:00 2001

From: Fangrui Song <[email protected]>
Date: Fri, 15 Nov 2019 16:27:47 -0800
Subject: [PATCH] Fix incorrect integer->float conversions caught by clang -Wimplicit-int-float-conversion
To: [email protected]

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

Quoted 13 lines

Review void qmp_migrate_set_downtime(double value, Error **errp) · line 2040

Quoted 10 lines

+++ b/migration/migration.c

(quoted diff) +1 -2
@@ -2035,11 +2035,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
    }
    value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
    MigrateSetParameters p = {
        .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)value,
    };

The explicit cast looks odd without a comment (generally, we try to avoid casts, so a comment such as /* explicit cast to silence compiler */ can be useful)

downtime_limit is an int64_t while value is a double.

There is a diagnostic (-Wfloat-conversion, included by -Wconversion)

warning: conversion from ‘double’ to ‘int64_t’ {aka ‘long int’} may change value [-Wfloat-conversion]

but it is not enabled by -Wall or -Wextra.

I am not familiar with qemu coding style, but I strongly feel it is a good thing to add an explicit cast. If it does not fit the style, I hope a maintainer can delete that for me.

Quoted 6 lines
Quoted 6 lines
(quoted diff) +0 -0
    qmp_migrate_set_parameters(&p, errp);
util/cutils.c +0 -0
index fd591cadf0..2b4484c015 100644
@@ -239,10 +239,10 @@ static int do_strtosz(const char *nptr, const char **end,
Quoted 15 lines

util/cutils.c static int do_strtosz(const char *nptr, const char **end, · line 243

Quoted 6 lines
util/cutils.c +1 -1
        goto out;
    }
    /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
+     * Values > nextafter(0x1p64, 0) overflow uint64_t after their trip
     * through double (53 bits of precision).

I thought we agreed on more text than just this (in particular, that the nextafter() call represents 2^64 rounded towards zero).

Quoted 6 lines
util/cutils.c +1 -1
     */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
        retval = -ERANGE;
        goto out;
    }

Sorry, I uploaded the wrong patch file. Attaching the correct one now.

From d533585df39083e88adc50b881a4be74125c837e Mon Sep 17 00:00:00 2001
From: Fangrui Song <[email protected]>
Date: Fri, 15 Nov 2019 16:27:47 -0800
Subject: [PATCH] Fix incorrect integer->float conversion caught by clang
-Wimplicit-int-float-conversion
To: [email protected]

The warning will be enabled by default in clang 10. It is not available for clang <= 9.

qemu/migration/migration.c:2038:24: error: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Werror,-Wimplicit-int-float-conversion]
...
qemu/util/cutils.c:245:23: error: implicit conversion from 'unsigned long' to 'double' changes value from 18446744073709550592 to 18446744073709551616 [-Werror,-Wimplicit-int-float-conversion]

---
 migration/migration.c | 3 +--
 util/cutils.c         | 8 +++++---
 2 files changed, 6 insertions(+), 5 deletions(-)
migration/migration.c +1 -2
index 354ad072fa..09b150663f 100644
@@ -2035,11 +2035,10 @@ void qmp_migrate_set_downtime(double value, Error **errp)
     }
 
     value *= 1000; /* Convert to milliseconds */
-    value = MAX(0, MIN(INT64_MAX, value));
 
     MigrateSetParameters p = {
         .has_downtime_limit = true,
-        .downtime_limit = value,
+        .downtime_limit = (int64_t)value,
     };
 
     qmp_migrate_set_parameters(&p, errp);
util/cutils.c +5 -3
index fd591cadf0..77acadc70a 100644
@@ -239,10 +239,12 @@ static int do_strtosz(const char *nptr, const char **end,
         goto out;
     }
     /*
-     * Values >= 0xfffffffffffffc00 overflow uint64_t after their trip
-     * through double (53 bits of precision).
+     * Values near UINT64_MAX overflow to 2**64 when converting to double
+     * precision.  Compare against the maximum representable double precision
+     * value below 2**64, computed as "the next value after 2**64 (0x1p64) in
+     * the direction of 0".
      */
-    if ((val * mul >= 0xfffffffffffffc00) || val < 0) {
+    if ((val * mul > nextafter(0x1p64, 0)) || val < 0) {
         retval = -ERANGE;
         goto out;
     }
Signature · 1 line
2.24.0
#10 lore raw

Re: [PATCH v2] Fix incorrect integer->float conversions caught by clang -Wimplicit-int-float-conversion

Quoted 9 lines · Fangrui Song
Quoted 8 lines · Eric Blake
Quoted 5 lines · Fangrui Song
Quoted 2 lines

Can we simply drop the offending line statement instead?

Fixed in the new patch.

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 2 lines

The first val * mul above this range is 0x1p64.  Rejecting it is correct, because it overflows yint64_t.

I am not subscribed, so apologize that this email may be off the thread.

(The binutils mailing list allows a user to download the raw email so I can still reply to a specific email, but this list does not provide such feature.)

There's <https://lists.gnu.org/archive/mbox/qemu-devel/>.

Quoted 9 lines
Quoted 7 lines

Actually, it's better to post a v2 patch as a new top-level thread, rather than buried as an attachment to a reply to v1, because our CI tooling doesn't see through the attachment (nor was it easy for me to reply to the v2 patch - I had to open the attachment to paste its text inline below...).

More patch submission hints at https://wiki.qemu.org/Contribute/SubmitAPatch

Retitled to [PATCH v2]

Good, such versioning is essential to avoid confusion. Next time, start a new top-level thread for v2. Our patch submission processing (automated as well as human) expects that.

I just did it for you: "[PATCH v3 0/2] Fix incorrect integer->float conversion caught by clang".

Many thanks for your fixes!