Skip to contentSkip to conversation

hw/s390x/ipl: Bail out if the network bootloader can not be found

PATCH Applied Thomas Huth 5 participants 8 messages Feb 27, 2018 – Mar 2, 2018

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hw/s390x/ipl: Bail out if the network bootloader can not be found

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
 hw/s390x/ipl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +3 -2
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
     netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, ipl->netboot_fw);
     if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",
+                   ipl->netboot_fw);
         goto unref_mr;
     }
 
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
     if (ipl->netboot) {
         if (load_netboot_image(&err) < 0) {
             error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);
+            exit(1);
         }
         ipl->iplb.ccw.netboot_start_addr = cpu_to_be64(ipl->start_addr);
     }
Signature · 1 line
1.8.3.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 32 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
 hw/s390x/ipl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +2 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
     netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, ipl->netboot_fw);
     if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",
+                   ipl->netboot_fw);
         goto unref_mr;
     }
 

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 423

Quoted 8 lines
hw/s390x/ipl.c +1 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
     if (ipl->netboot) {
         if (load_netboot_image(&err) < 0) {
             error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);
+            exit(1);
         }
         ipl->iplb.ccw.netboot_start_addr = cpu_to_be64(ipl->start_addr);
     }
Signature · 3 lines
Thanks,

David / dhildenb
#2 lore raw
In reply to Thomas Huth
Quoted 30 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
 hw/s390x/ipl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +1 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
     netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, ipl->netboot_fw);
     if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",> +                   ipl->netboot_fw);
         goto unref_mr;
     }
 

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 419

Quoted 4 lines
hw/s390x/ipl.c +0 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
     if (ipl->netboot) {
         if (load_netboot_image(&err) < 0) {
             error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);

Should we print something like 'exiting' or 'terminating' here, to make clear that the situation is terminal? Sometimes errors are reported and processing continues nonetheless.

Quoted 5 lines
hw/s390x/ipl.c +1 -0
+            exit(1);
         }
         ipl->iplb.ccw.netboot_start_addr = cpu_to_be64(ipl->start_addr);
     }
Signature · 2 lines
Regards,
 Viktor Mihajlovski
#3 lore raw
Quoted 30 lines · Viktor Mihajlovski
Quoted 29 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
 hw/s390x/ipl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +1 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
     netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, ipl->netboot_fw);
     if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",> +                   ipl->netboot_fw);
         goto unref_mr;
     }
 
Quoted 7 lines

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 419

Quoted 5 lines
hw/s390x/ipl.c +0 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
     if (ipl->netboot) {
         if (load_netboot_image(&err) < 0) {
             error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);

Should we print something like 'exiting' or 'terminating' here, to make clear that the situation is terminal?

No, we normally don't print out "exiting" or "terminating" in such a case. For example:

hw/alpha/dp264.c:    error_report("could not load palcode '%s'", palcode_filename);
hw/alpha/dp264.c:    exit(1);
hw/arm/boot.c:       error_report("failed to load \"%s\"", image_name);
hw/arm/boot.c:       exit(1);

hw/mips/mips_jazz.c: error_report("Could not load MIPS bios '%s'", bios_name); hw/mips/mips_jazz.c: exit(1);

hw/ppc/spapr.c:      error_report("Could not load LPAR rtas '%s'", filename);
hw/ppc/spapr.c:      exit(1);

etc.

Thomas

#4 lore raw
In reply to Viktor Mihajlovski
Quoted 30 lines · Viktor Mihajlovski
Quoted 29 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
  hw/s390x/ipl.c | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +1 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
      netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, ipl->netboot_fw);
      if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",> +                   ipl->netboot_fw);
          goto unref_mr;
      }
 
Quoted 8 lines

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 419

Quoted 5 lines
hw/s390x/ipl.c +0 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
      if (ipl->netboot) {
          if (load_netboot_image(&err) < 0) {
              error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);

Should we print something like 'exiting' or 'terminating' here, to make clear that the situation is terminal? Sometimes errors are reported and processing continues nonetheless.

I had to go through my old notes to see why I didn't just exit when I wrote it, and the reason was so we could put the guest in wait state so we can do some diagnostics....

Do we want to change this behavior?

Quoted 6 lines
Quoted 5 lines
hw/s390x/ipl.c +1 -0
+            exit(1);
          }
          ipl->iplb.ccw.netboot_start_addr = cpu_to_be64(ipl->start_addr);
      }
#5 lore raw
Quoted 45 lines · Farhan Ali
Quoted 32 lines · Viktor Mihajlovski
Quoted 31 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
  hw/s390x/ipl.c | 5 +++--

  1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +2 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
       netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS,
 ipl->netboot_fw);
       if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",>
+                   ipl->netboot_fw);
           goto unref_mr;
       }
Quoted 8 lines

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 419

Quoted 5 lines
hw/s390x/ipl.c +0 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
       if (ipl->netboot) {
           if (load_netboot_image(&err) < 0) {
               error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);

Should we print something like 'exiting' or 'terminating' here, to make clear that the situation is terminal? Sometimes errors are reported and processing continues nonetheless.

Quoted 5 lines

I had to go through my old notes to see why I didn't just exit when I wrote it, and the reason was so we could put the guest in wait state so we can do some diagnostics....

Do we want to change this behavior?

That intended behavior obviously does not work, since the guest is started anyway here.

And in this case, it does not make much sense to put the guest into wait state, since the problem is simply that a firmware image could not be loaded by QEMU, i.e. it's a problem in the host, not in the guest. Or which kind of guest diagnostics would you expect here? ... Putting the guest into stopped state only makes sense if the guest crashed / panicked, so you could analyze the reason for the crash.

Thomas

Reviewed-by
#6 lore raw
Quoted 55 lines · Thomas Huth
Quoted 45 lines · Farhan Ali
Quoted 32 lines · Viktor Mihajlovski
Quoted 31 lines · Thomas Huth

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

---
   hw/s390x/ipl.c | 5 +++--
   1 file changed, 3 insertions(+), 2 deletions(-)
hw/s390x/ipl.c +2 -1
index 0d06fc1..ff8308e 100644
@@ -322,7 +322,8 @@ static int load_netboot_image(Error **errp)
 
       netboot_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS,
 ipl->netboot_fw);
       if (netboot_filename == NULL) {
-        error_setg(errp, "Could not find network bootloader");
+        error_setg(errp, "Could not find network bootloader '%s'",>
+                   ipl->netboot_fw);
           goto unref_mr;
       }
Quoted 8 lines

hw/s390x/ipl.c void s390_ipl_prepare_cpu(S390CPU *cpu) · line 419

Quoted 5 lines
hw/s390x/ipl.c +0 -1
@@ -416,7 +417,7 @@ void s390_ipl_prepare_cpu(S390CPU *cpu)
       if (ipl->netboot) {
           if (load_netboot_image(&err) < 0) {
               error_report_err(err);
-            vm_stop(RUN_STATE_INTERNAL_ERROR);

Should we print something like 'exiting' or 'terminating' here, to make clear that the situation is terminal? Sometimes errors are reported and processing continues nonetheless.

Quoted 5 lines

I had to go through my old notes to see why I didn't just exit when I wrote it, and the reason was so we could put the guest in wait state so we can do some diagnostics....

Do we want to change this behavior?

That intended behavior obviously does not work, since the guest is started anyway here.

Quoted 9 lines

And in this case, it does not make much sense to put the guest into wait state, since the problem is simply that a firmware image could not be loaded by QEMU, i.e. it's a problem in the host, not in the guest. Or which kind of guest diagnostics would you expect here? ... Putting the guest into stopped state only makes sense if the guest crashed / panicked, so you could analyze the reason for the crash.

Thomas

You are right and this would be the right thing to do.

#7 lore raw
In reply to Thomas Huth
Quoted 11 lines · On Tue, 27 Feb 2018 11:05:13 +0100 Thomas Huth <[email protected]> wrote:

If QEMU fails to load 's390-netboot.img', the guest firmware currently loops forever and just floods the console with "Network boot device detected" messages. The code in ipl.c apparently already tried to stop the VM with vm_stop() in this case, but this is in vain since the run state is later reset due to a call to vm_start() from vl.c again. To avoid the ugly firmware loop, let's simply exit QEMU directly instead since it just does not make sense to continue if the required firmware image can not be loaded. While we're at it, also add the file name of the netboot binary to the error message, so that the user has a better hint about what is missing.

Quoted 4 lines
---
 hw/s390x/ipl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

Thanks, applied.