Skip to contentSkip to conversation

misc bugfixes

PATCH In review Gerd Hoffmann 5 participants 17 messages Dec 10, 2009 – Dec 11, 2009

lore
Share & tools
lore
#0 lore raw

cover misc bugfixes

Hi,

A bunch of fixes for various bugs, mostly hotplug stuff found by Daniel.

please apply,
Gerd

#1 lore raw

PATCH 1/4 pci: don't abort() when trying to hotplug with acpi off.

The PCI bus on x86 requires ACPI for hotplug support, thus disbling ACPI also disables hotplug for the PCI bus. This patch makes qemu check whenever the PCI bus in question can handle hotplug before trying to add devices. This is needed because qdev will abort() on any attempt to hotplug devices into a non-hotpluggable bus.

---
 hw/pci-hotplug.c |   17 ++++++++++++++++-
 1 files changed, 16 insertions(+), 1 deletions(-)
hw/pci-hotplug.c +16 -1
index 081d6d1..7e5c51d 100644
@@ -40,7 +40,18 @@ static PCIDevice *qemu_pci_hot_add_nic(Monitor *mon,
                                        const char *opts_str)
 {
     QemuOpts *opts;
-    int ret;
+    PCIBus *bus;
+    int ret, devfn;
+
+    bus = pci_get_bus_devfn(&devfn, devaddr);
+    if (!bus) {
+        monitor_printf(mon, "Invalid PCI device address %s\n", devaddr);
+        return NULL;
+    }
+    if (!((BusState*)bus)->allow_hotplug) {
+        monitor_printf(mon, "PCI bus doesn't support hotplug\n");
+        return NULL;
+    }
 
     opts = qemu_opts_parse(&qemu_net_opts, opts_str ? opts_str : "", NULL);
     if (!opts) {
@@ -179,6 +190,10 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
         monitor_printf(mon, "Invalid PCI device address %s\n", devaddr);
         return NULL;
     }
+    if (!((BusState*)bus)->allow_hotplug) {
+        monitor_printf(mon, "PCI bus doesn't support hotplug\n");
+        return NULL;
+    }
 
     switch (type) {
     case IF_SCSI:
Signature · 1 line
1.6.5.2
#2 lore raw

PATCH 2/4 pci: don't hw_error() when no slot is available.

In reply to Gerd Hoffmann

Current PCI code will simply hw_error() and thus abort in case no free PCI slot is available or the requested PCI slot is already in use by another device. For the hotplug case this behavior is not acceptable. This patch makes qemu pass up the error properly, so the calling code can decide whenever it wants to exit with an error (on startup) or whenever it wants to continue (hotplug).

---
 hw/pci.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)
hw/pci.c +9 -2
index 4f662b7..404eead 100644
@@ -580,11 +580,13 @@ static PCIDevice *do_pci_register_device(PCIDevice *pci_dev, PCIBus *bus,
             if (!bus->devices[devfn])
                 goto found;
         }
-        hw_error("PCI: no devfn available for %s, all in use\n", name);
+        qemu_error("PCI: no devfn available for %s, all in use\n", name);
+        return NULL;
     found: ;
     } else if (bus->devices[devfn]) {
-        hw_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
+        qemu_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
                  name, bus->devices[devfn]->name);
+        return NULL;
     }
     pci_dev->bus = bus;
     pci_dev->devfn = devfn;
@@ -625,6 +627,9 @@ PCIDevice *pci_register_device(PCIBus *bus, const char *name,
     pci_dev = do_pci_register_device(pci_dev, bus, name, devfn,
                                      config_read, config_write,
                                      PCI_HEADER_TYPE_NORMAL);
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }
     return pci_dev;
 }
 static target_phys_addr_t pci_to_cpu_addr(target_phys_addr_t addr)
@@ -1376,6 +1381,8 @@ static int pci_qdev_init(DeviceState *qdev, DeviceInfo *base)
     pci_dev = do_pci_register_device(pci_dev, bus, base->name, devfn,
                                      info->config_read, info->config_write,
                                      info->header_type);
+    if (pci_dev == NULL)
+        return -1;
     rc = info->init(pci_dev);
     if (rc != 0)
         return rc;
Signature · 1 line
1.6.5.2
#3 lore raw

[Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 8 lines · Gerd Hoffmann

Current PCI code will simply hw_error() and thus abort in case no free PCI slot is available or the requested PCI slot is already in use by another device. For the hotplug case this behavior is not acceptable. This patch makes qemu pass up the error properly, so the calling code can decide whenever it wants to exit with an error (on startup) or whenever it wants to continue (hotplug).

Good stuff. However

Quoted 25 lines
---
 hw/pci.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)
hw/pci.c +4 -2
index 4f662b7..404eead 100644
@@ -580,11 +580,13 @@ static PCIDevice *do_pci_register_device(PCIDevice *pci_dev, PCIBus *bus,
             if (!bus->devices[devfn])
                 goto found;
         }
-        hw_error("PCI: no devfn available for %s, all in use\n", name);
+        qemu_error("PCI: no devfn available for %s, all in use\n", name);
+        return NULL;
     found: ;
     } else if (bus->devices[devfn]) {
-        hw_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
+        qemu_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
                  name, bus->devices[devfn]->name);
+        return NULL;
     }
     pci_dev->bus = bus;
     pci_dev->devfn = devfn;

hw/pci.c PCIDevice *pci_register_device(PCIBus *bus, const char *name, · line 632

Quoted 6 lines
hw/pci.c +3 -0
@@ -625,6 +627,9 @@ PCIDevice *pci_register_device(PCIBus *bus, const char *name,
     pci_dev = do_pci_register_device(pci_dev, bus, name, devfn,
                                      config_read, config_write,
                                      PCI_HEADER_TYPE_NORMAL);
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }

Can you please use !pci_dev for these checks?

hw/pci.c static int pci_qdev_init(DeviceState *qdev, DeviceInfo *base) · line 1385

Quoted 9 lines
hw/pci.c +2 -0
     return pci_dev;
 }
 static target_phys_addr_t pci_to_cpu_addr(target_phys_addr_t addr)
@@ -1376,6 +1381,8 @@ static int pci_qdev_init(DeviceState *qdev, DeviceInfo *base)
     pci_dev = do_pci_register_device(pci_dev, bus, base->name, devfn,
                                      info->config_read, info->config_write,
                                      info->header_type);
+    if (pci_dev == NULL)
+        return -1;

And here too.

Quoted 7 lines
hw/pci.c +0 -0
     rc = info->init(pci_dev);
     if (rc != 0)
         return rc;
Signature · 1 line
1.6.5.2
#4 lore raw

[Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Hi,

Quoted 5 lines

hw/pci.c PCIDevice *pci_register_device(PCIBus *bus, const char *name, · line 632

Quoted 3 lines
hw/pci.c +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

#5 lore raw

[Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines
Quoted 3 lines
hw/pci.c +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }
Quoted 8 lines
Quoted 1 line

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

C is a terse language. !x is a standard idiom. Let's use it.

Signature · 1 line
MST
#6 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 13 lines · Michael S. Tsirkin
Quoted 12 lines · Gerd Hoffmann

Hi,

Quoted 5 lines

hw/pci.c PCIDevice *pci_register_device(PCIBus *bus, const char *name, · line 632

Quoted 3 lines
hw/pci.c +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

Quoted 9 lines
Quoted 2 lines

cheers,
Gerd

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

C is a terse language. !x is a standard idiom. Let's use it.

! is for boolean. pci_dev is not boolean.

--
Gleb.

#7 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 17 lines · Gleb Natapov
Quoted 16 lines · Michael S. Tsirkin
Quoted 14 lines · Gerd Hoffmann

Hi,

Quoted 5 lines

hw/pci.c PCIDevice *pci_register_device(PCIBus *bus, const char *name, · line 632

Quoted 3 lines
hw/pci.c +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

Quoted 7 lines
Quoted 6 lines

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

C is a terse language. !x is a standard idiom. Let's use it.

! is for boolean.

In which language? Not in C.

Quoted 3 lines

pci_dev is not boolean.
--
Gleb.

#8 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 22 lines · Michael S. Tsirkin
Quoted 21 lines · Gleb Natapov
Quoted 20 lines · Michael S. Tsirkin
Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines
Quoted 3 lines
hw/pci.c +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }
Quoted 8 lines
Quoted 1 line

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

Quoted 6 lines
Quoted 3 lines
Quoted 2 lines

C is a terse language. !x is a standard idiom. Let's use it.

! is for boolean.

In which language? Not in C.

In boolean. My point is no need to force your style on everyone if it is not in coding style document. == often much more readable then !.

Quoted 3 lines
Quoted 3 lines

pci_dev is not boolean.
--
Gleb.

--
Gleb.

#9 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 23 lines · Gleb Natapov
Quoted 22 lines · Michael S. Tsirkin
Quoted 21 lines · Gleb Natapov
Quoted 20 lines · Michael S. Tsirkin
Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines
Quoted 3 lines
(quoted diff) +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }
Quoted 8 lines
Quoted 1 line

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

Quoted 7 lines
Quoted 6 lines
Quoted 3 lines
Quoted 2 lines

C is a terse language. !x is a standard idiom. Let's use it.

! is for boolean.

In which language? Not in C.

In boolean. My point is no need to force your style on everyone

I am not forcing style on anyone - how can I do this? I hacked on pci and by now there's no == NULL in pci.c anywhere ;). I will be happier if none is added now.

Quoted 1 line

if it is not in coding style document.

Heh, we don't need to document everything.

Quoted 1 line

== often much more readable then !.

Matter of taste probably.

Quoted 6 lines
Quoted 3 lines
Quoted 3 lines

pci_dev is not boolean.
--
Gleb.

--
Gleb.

#10 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

Quoted 36 lines · Michael S. Tsirkin
Quoted 23 lines · Gleb Natapov
Quoted 22 lines · Michael S. Tsirkin
Quoted 21 lines · Gleb Natapov
Quoted 20 lines · Michael S. Tsirkin
Quoted 6 lines · Gerd Hoffmann

Hi,

Quoted 4 lines
Quoted 3 lines
(quoted diff) +3 -0
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }
Quoted 8 lines
Quoted 1 line

Can you please use !pci_dev for these checks?

Why? IMHO the code is more readable that way. It is easy to miss a single '!' character when reading the code, so I tend to write such tests in a more verbose fashion.

cheers,
Gerd

Reader has limited short term memory. Don't fill it up with irrelevant detail. In places where it might be confusing, a comment might be appropriate, this is not one of them.

Quoted 7 lines
Quoted 6 lines
Quoted 3 lines
Quoted 2 lines

C is a terse language. !x is a standard idiom. Let's use it.

! is for boolean.

In which language? Not in C.

In boolean. My point is no need to force your style on everyone

I am not forcing style on anyone - how can I do this? I hacked on pci and by now there's no == NULL in pci.c anywhere ;). I will be happier if none is added now.

Quoted 8 lines
Quoted 1 line

if it is not in coding style document.

Heh, we don't need to document everything.

Quoted 1 line

== often much more readable then !.

Matter of taste probably.

Exactly! (Or should I say "Exactly=="?)

--
Gleb.

#11 lore raw

Re: [Qemu-devel] Re: [FOR 0.12 PATCH 2/4] pci: don't hw_error() when no slot is available.

In reply to Michael S. Tsirkin
Quoted 47 lines · Michael S. Tsirkin
Quoted 8 lines · Gerd Hoffmann

Current PCI code will simply hw_error() and thus abort in case no free PCI slot is available or the requested PCI slot is already in use by another device. For the hotplug case this behavior is not acceptable. This patch makes qemu pass up the error properly, so the calling code can decide whenever it wants to exit with an error (on startup) or whenever it wants to continue (hotplug).

Good stuff. However

Quoted 25 lines

---
hw/pci.c | 11 +++++++++--

1 files changed, 9 insertions(+), 2 deletions(-)
hw/pci.c +4 -2
index 4f662b7..404eead 100644
@@ -580,11 +580,13 @@ static PCIDevice *do_pci_register_device(PCIDevice *pci_dev, PCIBus *bus,
            if (!bus->devices[devfn])
                goto found;
        }
-        hw_error("PCI: no devfn available for %s, all in use\n", name);
+        qemu_error("PCI: no devfn available for %s, all in use\n", name);
+        return NULL;
    found: ;
    } else if (bus->devices[devfn]) {
-        hw_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
+        qemu_error("PCI: devfn %d not available for %s, in use by %s\n", devfn,
                 name, bus->devices[devfn]->name);
+        return NULL;
    }
    pci_dev->bus = bus;
    pci_dev->devfn = devfn;

hw/pci.c PCIDevice *pci_register_device(PCIBus *bus, const char *name, · line 632

Quoted 6 lines
hw/pci.c +3 -0
@@ -625,6 +627,9 @@ PCIDevice *pci_register_device(PCIBus *bus, const char *name,
    pci_dev = do_pci_register_device(pci_dev, bus, name, devfn,
                                     config_read, config_write,
                                     PCI_HEADER_TYPE_NORMAL);
+    if (pci_dev == NULL) {
+        hw_error("PCI: can't register device\n");
+    }

Can you please use !pci_dev for these checks?

Quoted 11 lines

hw/pci.c static int pci_qdev_init(DeviceState *qdev, DeviceInfo *base) · line 1385

Quoted 9 lines
hw/pci.c +2 -0
    return pci_dev;
 }
 static target_phys_addr_t pci_to_cpu_addr(target_phys_addr_t addr)
@@ -1376,6 +1381,8 @@ static int pci_qdev_init(DeviceState *qdev, DeviceInfo *base)
    pci_dev = do_pci_register_device(pci_dev, bus, base->name, devfn,
                                     info->config_read, info->config_write,
                                     info->header_type);
+    if (pci_dev == NULL)
+        return -1;

And here too.

OMG! Broken coding style!

:)

Alex

#12 lore raw

PATCH 3/4 scsi: fix drive hotplug.

In reply to Gerd Hoffmann

This patch fills the DriveInfo->unit after hotplugging a scsi disk. It makes a difference when auto-assigning a scsi id, where unit was left filled with '-1' instead of the actual scsi id.

With this patch applied the the drive naming logic in drive_init() works as good as it did in previous releases. Which means it works fine with a single scsi bus.

---
 hw/pci-hotplug.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
hw/pci-hotplug.c +1 -0
index 7e5c51d..9e8e6ed 100644
@@ -93,6 +93,7 @@ static int scsi_hot_add(DeviceState *adapter, DriveInfo *dinfo, int printinfo)
      */
     dinfo->unit = qemu_opt_get_number(dinfo->opts, "unit", -1);
     scsidev = scsi_bus_legacy_add_drive(scsibus, dinfo, dinfo->unit);
+    dinfo->unit = scsidev->id;
 
     if (printinfo)
         qemu_error("OK bus %d, unit %d\n", scsibus->busnr, scsidev->id);
Signature · 1 line
1.6.5.2
#13 lore raw

PATCH 4/4 QemuOpts: allow larger option values.

In reply to Gerd Hoffmann

Use case: loooooooooooooooooong file names for -drive file=...

---
 qemu-option.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
qemu-option.c +2 -2
index b009109..24392fc 100644
@@ -705,7 +705,7 @@ int qemu_opts_print(QemuOpts *opts, void *dummy)
 
 int qemu_opts_do_parse(QemuOpts *opts, const char *params, const char *firstname)
 {
-    char option[128], value[128];
+    char option[128], value[1024];
     const char *p,*pe,*pc;
 
     for (p = params; *p != '\0'; p++) {
@@ -751,7 +751,7 @@ int qemu_opts_do_parse(QemuOpts *opts, const char *params, const char *firstname
 
 QemuOpts *qemu_opts_parse(QemuOptsList *list, const char *params, const char *firstname)
 {
-    char value[128], *id = NULL;
+    char value[1024], *id = NULL;
     const char *p;
     QemuOpts *opts;
 
Signature · 1 line
1.6.5.2
#14 lore raw
In reply to Gerd Hoffmann
Quoted 7 lines · Gerd Hoffmann

Hi,

A bunch of fixes for various bugs, mostly hotplug stuff found by Daniel.

please apply,
Gerd

I am kind of confused by the FOR 0.12 prefix - these patches are not on master yet, are they? Should patches land on master before being backported to 0.12?

Signature · 1 line
MST
#15 lore raw
Quoted 11 lines · Michael S. Tsirkin
Quoted 7 lines · Gerd Hoffmann
   Hi,

A bunch of fixes for various bugs, mostly hotplug stuff found by Daniel.

please apply,
Gerd

I am kind of confused by the FOR 0.12 prefix - these patches are not on master yet, are they?

No.

Quoted 2 lines

Should patches land on master before being backported to 0.12?

Anthony asked to tag patches this way if they should be merged for the 0.12 release. So, after branching, they should go to both master and 0.12 I think. Do we have a 0.12 branch already?

cheers,
Gerd

#16 lore raw
In reply to Michael S. Tsirkin
Quoted 11 lines · Michael S. Tsirkin
Quoted 8 lines · Gerd Hoffmann

Hi,

A bunch of fixes for various bugs, mostly hotplug stuff found by Daniel.

please apply,
Gerd

Quoted 5 lines

I am kind of confused by the FOR 0.12 prefix - these patches are not on master yet, are they? Should patches land on master before being backported to 0.12?

Right now, there's almost no difference between master and stable-0.12. I always commit to master first and then cherry pick.

FOR 0.12 helps me know that I need to cherry pick from master into stable-0.12.

Regards,

Anthony Liguori