Skip to contentSkip to conversation

fw_cfg_reboot: ensure reboot_time is nonegative

PATCH In review Li Qiang 5 participants 9 messages Oct 24, 2018 – Nov 1, 2018

lore
Share & tools
lore

Li Qiang

author
#0 lore raw

PATCH 1/1 fw_cfg_reboot: ensure reboot_time is nonegative

This can avoid setting a negative value to
etc/boot-fail-wait.

---
 hw/nvram/fw_cfg.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
hw/nvram/fw_cfg.c +10 -5
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
             reboot_timeout = strtol(p, &p, 10);
         }
     }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to 65535.");
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"
+                         "force it to 65535.");
+            reboot_timeout = 0xffff;
+        }
+        fw_cfg_add_file(s, "etc/boot-fail-wait",
+                        g_memdup(&reboot_timeout, 4), 4);
     }
-    fw_cfg_add_file(s, "etc/boot-fail-wait", g_memdup(&reboot_timeout, 4), 4);
 }
 
 static void fw_cfg_write(FWCfgState *s, uint8_t value)
Signature · 1 line
1.8.3.1
#1 lore raw
Quoted 26 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

---
 hw/nvram/fw_cfg.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
hw/nvram/fw_cfg.c +5 -4
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
             reboot_timeout = strtol(p, &p, 10);
         }
     }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to 65535.");
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 215

Quoted 10 lines
hw/nvram/fw_cfg.c +5 -1
+                         "force it to 65535.");
+            reboot_timeout = 0xffff;
+        }
+        fw_cfg_add_file(s, "etc/boot-fail-wait",
+                        g_memdup(&reboot_timeout, 4), 4);
     }
-    fw_cfg_add_file(s, "etc/boot-fail-wait", g_memdup(&reboot_timeout, 4), 4);
 }
 
 static void fw_cfg_write(FWCfgState *s, uint8_t value)

I don't feel strongly about fixing this issue.

However, if we decide to fix it, we should start with the bare-bones strtol() call, visible at the top of the context. I'm not up-to-date on what's the best QEMU helper function for this, but I seem to remember it checks for trailing garbage, and perhaps even for range. Maybe we should even use a different (better) option parsing facility thatn qemu_opt_get(). Adding Eric and Markus.

Also, I would suggest forcing negative values (that were explicitly specified) to some sensible positive default, such as 5 seconds or so.

Thanks
Laszlo

#2 lore raw

Hi,

Quoted 3 lines · Laszlo Ersek
Quoted 2 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

Li Qiang, can you add a qtest for this?

Quoted 37 lines
Quoted 24 lines
---
  hw/nvram/fw_cfg.c | 15 ++++++++++-----
  1 file changed, 10 insertions(+), 5 deletions(-)
hw/nvram/fw_cfg.c +5 -4
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
              reboot_timeout = strtol(p, &p, 10);
          }
      }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to 65535.");
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 215

Quoted 10 lines
hw/nvram/fw_cfg.c +5 -1
+                         "force it to 65535.");
+            reboot_timeout = 0xffff;
+        }
+        fw_cfg_add_file(s, "etc/boot-fail-wait",
+                        g_memdup(&reboot_timeout, 4), 4);
      }
-    fw_cfg_add_file(s, "etc/boot-fail-wait", g_memdup(&reboot_timeout, 4), 4);
  }
 
  static void fw_cfg_write(FWCfgState *s, uint8_t value)

I don't feel strongly about fixing this issue.

Quoted 4 lines

However, if we decide to fix it, we should start with the bare-bones strtol() call, visible at the top of the context. I'm not up-to-date on what's the best QEMU helper function for this, but I seem to remember it checks for trailing garbage, and perhaps even for range. Maybe we should

Are you suggesting qemu_strtoul()? I agree this would be cleaner.

Quoted 9 lines

even use a different (better) option parsing facility thatn qemu_opt_get(). Adding Eric and Markus.

Also, I would suggest forcing negative values (that were explicitly specified) to some sensible positive default, such as 5 seconds or so.

Thanks
Laszlo

author
#3 lore raw

Hello Laszlo and Philippe ,

Thanks for your review,

Quoted 10 lines · Philippe Mathieu-Daudé <[email protected]> 于2018年10月25日周四 上午6:56写道:

Hi,

Quoted 3 lines · Laszlo Ersek
Quoted 2 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

Li Qiang, can you add a qtest for this?

I will try to write one.
Is it ok to write it without this patch, as the test will be not passed?

Thanks,
Li Qiang

Quoted 37 lines
Quoted 6 lines
Quoted 6 lines
---
  hw/nvram/fw_cfg.c | 15 ++++++++++-----
  1 file changed, 10 insertions(+), 5 deletions(-)
Quoted 11 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 204

Quoted 11 lines
hw/nvram/fw_cfg.c +0 -3
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
              reboot_timeout = strtol(p, &p, 10);
          }
      }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to

65535.");

Quoted 5 lines
Quoted 5 lines
hw/nvram/fw_cfg.c +4 -1
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
Quoted 8 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s)

Quoted 8 lines
hw/nvram/fw_cfg.c +6 -1
+            error_report("reboot timeout is larger than 65535,"
+                         "force it to 65535.");
+            reboot_timeout = 0xffff;
+        }
+        fw_cfg_add_file(s, "etc/boot-fail-wait",
+                        g_memdup(&reboot_timeout, 4), 4);
      }
-    fw_cfg_add_file(s, "etc/boot-fail-wait", g_memdup(&reboot_timeout,

4), 4);

Quoted 5 lines
Quoted 4 lines
hw/nvram/fw_cfg.c +0 -0
  }
 
  static void fw_cfg_write(FWCfgState *s, uint8_t value)
Quoted 10 lines
Quoted 6 lines

I don't feel strongly about fixing this issue.

However, if we decide to fix it, we should start with the bare-bones strtol() call, visible at the top of the context. I'm not up-to-date on what's the best QEMU helper function for this, but I seem to remember it checks for trailing garbage, and perhaps even for range. Maybe we should

Are you suggesting qemu_strtoul()? I agree this would be cleaner.

I reference the 'boot_splash_time' in fw_cfg_bootsplash. We can also change there.

Quoted 10 lines
Quoted 9 lines

even use a different (better) option parsing facility thatn qemu_opt_get(). Adding Eric and Markus.

Also, I would suggest forcing negative values (that were explicitly specified) to some sensible positive default, such as 5 seconds or so.

Thanks
Laszlo

#4 lore raw
Quoted 18 lines · Li Qiang

Hello Laszlo and Philippe ,

Thanks for your review,

Philippe Mathieu-Daudé <[email protected] <mailto:[email protected]>> 于 2018年10月25日周四 上午6:56写道:

    Hi,

    On 24/10/18 13:35, Laszlo Ersek wrote:
     > On 10/24/18 09:11, Li Qiang wrote:
     >> This can avoid setting a negative value to
     >> etc/boot-fail-wait.

    Li Qiang, can you add a qtest for this?
Quoted 2 lines

I will try to write one.
Is it ok to write it without this patch, as the test will be not passed?

It is OK to add the test after the fix, else the test would fail. But now your test protect from future regressions.

Quoted 70 lines

Thanks,
Li Qiang

     >>
     >> Signed-off-by: Li Qiang <[email protected] <mailto:[email protected]>>
     >> ---
     >>   hw/nvram/fw_cfg.c | 15 ++++++++++-----
     >>   1 file changed, 10 insertions(+), 5 deletions(-)
     >>
     >> diff --git a/hw/nvram/fw_cfg.c b/hw/nvram/fw_cfg.c
     >> index f4a52d8..276dcb1 100644
     >> --- a/hw/nvram/fw_cfg.c
     >> +++ b/hw/nvram/fw_cfg.c
     >> @@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
     >>               reboot_timeout = strtol(p, &p, 10);
     >>           }
     >>       }
     >> -    /* validate the input */
     >> -    if (reboot_timeout > 0xffff) {
     >> -        error_report("reboot timeout is larger than 65535,
    force it to 65535.");
     >> -        reboot_timeout = 0xffff;
     >> +
     >> +    if (reboot_timeout >= 0) {
     >> +        /* validate the input */
     >> +        if (reboot_timeout > 0xffff) {
     >> +            error_report("reboot timeout is larger than 65535,"
     >> +                         "force it to 65535.");
     >> +            reboot_timeout = 0xffff;
     >> +        }
     >> +        fw_cfg_add_file(s, "etc/boot-fail-wait",
     >> +                        g_memdup(&reboot_timeout, 4), 4);
     >>       }
     >> -    fw_cfg_add_file(s, "etc/boot-fail-wait",
    g_memdup(&reboot_timeout, 4), 4);
     >>   }
     >>
     >>   static void fw_cfg_write(FWCfgState *s, uint8_t value)
     >>
     >
     > I don't feel strongly about fixing this issue.
     >
     > However, if we decide to fix it, we should start with the bare-bones
     > strtol() call, visible at the top of the context. I'm not
    up-to-date on
     > what's the best QEMU helper function for this, but I seem to
    remember it
     > checks for trailing garbage, and perhaps even for range. Maybe we
    should

    Are you suggesting qemu_strtoul()? I agree this would be cleaner.

I reference the 'boot_splash_time' in fw_cfg_bootsplash. We can also change there.

> even use a different (better) option parsing facility thatn > qemu_opt_get(). Adding Eric and Markus.
>
> Also, I would suggest forcing negative values (that were explicitly > specified) to some sensible positive default, such as 5 seconds
or so.
>
> Thanks
> Laszlo
>

author
#5 lore raw
Quoted 17 lines · Philippe Mathieu-Daudé <[email protected]> 于2018年10月30日周二 上午7:50写道:
Quoted 16 lines · Li Qiang

Hello Laszlo and Philippe ,

Thanks for your review,

Philippe Mathieu-Daudé <[email protected] <mailto:[email protected]>> 于 2018年10月25日周四 上午6:56写道:

    Hi,

    On 24/10/18 13:35, Laszlo Ersek wrote:
     > On 10/24/18 09:11, Li Qiang wrote:
     >> This can avoid setting a negative value to
     >> etc/boot-fail-wait.
Quoted 10 lines
Quoted 5 lines

Li Qiang, can you add a qtest for this?

I will try to write one.
Is it ok to write it without this patch, as the test will be not passed?

It is OK to add the test after the fix, else the test would fail. But now your test protect from future regressions.

Ok, I think this fix is not that bad, this is the same as 'boot_splash_time'.
As for Laszlo's consideration, we think we can send another patch to
drop strtol in both 'boot_splash_time' and this 'reboot-timeout'.

Thanks,
Li Qiang

Quoted 74 lines
Quoted 7 lines

Thanks,
Li Qiang

>>
>> Signed-off-by: Li Qiang <[email protected] <mailto:

[email protected]>>

Quoted 30 lines
     >> ---
     >>   hw/nvram/fw_cfg.c | 15 ++++++++++-----
     >>   1 file changed, 10 insertions(+), 5 deletions(-)
     >>
     >> diff --git a/hw/nvram/fw_cfg.c b/hw/nvram/fw_cfg.c
     >> index f4a52d8..276dcb1 100644
     >> --- a/hw/nvram/fw_cfg.c
     >> +++ b/hw/nvram/fw_cfg.c
     >> @@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
     >>               reboot_timeout = strtol(p, &p, 10);
     >>           }
     >>       }
     >> -    /* validate the input */
     >> -    if (reboot_timeout > 0xffff) {
     >> -        error_report("reboot timeout is larger than 65535,
    force it to 65535.");
     >> -        reboot_timeout = 0xffff;
     >> +
     >> +    if (reboot_timeout >= 0) {
     >> +        /* validate the input */
     >> +        if (reboot_timeout > 0xffff) {
     >> +            error_report("reboot timeout is larger than 65535,"
     >> +                         "force it to 65535.");
     >> +            reboot_timeout = 0xffff;
     >> +        }
     >> +        fw_cfg_add_file(s, "etc/boot-fail-wait",
     >> +                        g_memdup(&reboot_timeout, 4), 4);
     >>       }
     >> -    fw_cfg_add_file(s, "etc/boot-fail-wait",
    g_memdup(&reboot_timeout, 4), 4);
Quoted 8 lines
     >>   }
     >>
     >>   static void fw_cfg_write(FWCfgState *s, uint8_t value)
     >>
     >
     > I don't feel strongly about fixing this issue.
     >
     > However, if we decide to fix it, we should start with the

bare-bones

Quoted 11 lines

> strtol() call, visible at the top of the context. I'm not
up-to-date on
> what's the best QEMU helper function for this, but I seem to
remember it
> checks for trailing garbage, and perhaps even for range. Maybe we
should

Are you suggesting qemu_strtoul()? I agree this would be cleaner.

Quoted 7 lines

I reference the 'boot_splash_time' in fw_cfg_bootsplash. We can also change there.

> even use a different (better) option parsing facility thatn > qemu_opt_get(). Adding Eric and Markus.
>
> Also, I would suggest forcing negative values (that were

explicitly

Quoted 7 lines
     > specified) to some sensible positive default, such as 5 seconds
    or so.
     >
     > Thanks
     > Laszlo
     >
#6 lore raw
In reply to Laszlo Ersek
Quoted 40 lines · Laszlo Ersek
Quoted 26 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

---
 hw/nvram/fw_cfg.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)
hw/nvram/fw_cfg.c +5 -4
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
             reboot_timeout = strtol(p, &p, 10);
         }
     }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to 65535.");
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 215

Quoted 10 lines
hw/nvram/fw_cfg.c +5 -1
+                         "force it to 65535.");
+            reboot_timeout = 0xffff;
+        }
+        fw_cfg_add_file(s, "etc/boot-fail-wait",
+                        g_memdup(&reboot_timeout, 4), 4);
     }
-    fw_cfg_add_file(s, "etc/boot-fail-wait", g_memdup(&reboot_timeout, 4), 4);
 }
 
 static void fw_cfg_write(FWCfgState *s, uint8_t value)

I don't feel strongly about fixing this issue.

Quoted 6 lines

However, if we decide to fix it, we should start with the bare-bones strtol() call, visible at the top of the context. I'm not up-to-date on what's the best QEMU helper function for this, but I seem to remember it checks for trailing garbage, and perhaps even for range. Maybe we should even use a different (better) option parsing facility thatn qemu_opt_get(). Adding Eric and Markus.

Whenever we qemu_opt_get() a string, then convert it to a number ourselves, we're doing it wrong. Can you use qemu_opt_get_number()?

When you have to convert string to integer, use the qemu_strtoX() that gives you the width and signedness you need. Or maybe one of the parse_uint*().

Quoted 5 lines

Also, I would suggest forcing negative values (that were explicitly specified) to some sensible positive default, such as 5 seconds or so.

Thanks
Laszlo

#7 lore raw
In reply to Laszlo Ersek
Quoted 9 lines · Laszlo Ersek
Quoted 8 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

---
  hw/nvram/fw_cfg.c | 15 ++++++++++-----
  1 file changed, 10 insertions(+), 5 deletions(-)
Quoted 6 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 199

Quoted 6 lines
hw/nvram/fw_cfg.c +0 -0
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
              reboot_timeout = strtol(p, &p, 10);

Looks like Markus handled the question about replacing strtol(), so I'll just point out one additional nit:

Quoted 5 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 204

Quoted 5 lines
hw/nvram/fw_cfg.c +0 -3
          }
      }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to 65535.");

Pre-existing, but now's as good a time as any to improve it:

Quoted 7 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 207

Quoted 7 lines
hw/nvram/fw_cfg.c +6 -1
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"
+                         "force it to 65535.");

error_report() callers generally do not end with trailing '.'

Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org
author
#8 lore raw

Thanks Eric and Markus's review,

I have sent out another patch.

I'm not sure what's the effect when the etc/reboot-timeout and etc/splash-time is 0
in seabios, so CC'd Gerd. Maybe it can be more simplicity.

Thanks,
Li Qiang

Quoted 44 lines · Eric Blake <[email protected]> 于2018年10月31日周三 下午11:55写道:
Quoted 9 lines · Laszlo Ersek
Quoted 8 lines · Li Qiang

This can avoid setting a negative value to
etc/boot-fail-wait.

---
  hw/nvram/fw_cfg.c | 15 ++++++++++-----
  1 file changed, 10 insertions(+), 5 deletions(-)
Quoted 6 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 199

Quoted 6 lines
hw/nvram/fw_cfg.c +0 -0
index f4a52d8..276dcb1 100644
@@ -199,12 +199,17 @@ static void fw_cfg_reboot(FWCfgState *s)
              reboot_timeout = strtol(p, &p, 10);

Looks like Markus handled the question about replacing strtol(), so I'll just point out one additional nit:

Quoted 5 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s)

Quoted 5 lines
hw/nvram/fw_cfg.c +0 -3
          }
      }
-    /* validate the input */
-    if (reboot_timeout > 0xffff) {
-        error_report("reboot timeout is larger than 65535, force it to

65535.");

Pre-existing, but now's as good a time as any to improve it:

Quoted 7 lines

hw/nvram/fw_cfg.c static void fw_cfg_reboot(FWCfgState *s) · line 207

Quoted 7 lines
hw/nvram/fw_cfg.c +6 -1
-        reboot_timeout = 0xffff;
+
+    if (reboot_timeout >= 0) {
+        /* validate the input */
+        if (reboot_timeout > 0xffff) {
+            error_report("reboot timeout is larger than 65535,"
+                         "force it to 65535.");

error_report() callers generally do not end with trailing '.'

--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org