Drivers: hv: vmbus: Export hv_vmbus_exists() and use it in pci-hyperv
Subsystems (4)
Dexuan Cui
maintainerPATCH 1/1 Drivers: hv: vmbus: Export hv_vmbus_exists() and use it in pci-hyperv
With commit f84b21da3624 ("PCI: hv: Don't load the driver for baremetal root partition"), the bare metal Linux root partition won't use the pci-hyperv driver, but when a Linux VM runs on the Linux root partition, pci-hyperv's module_init function init_hv_pci_drv() can still run, e.g. in the case of CONFIG_PCI_HYPERV=y, even if the VMBus driver is not used in such a VM (i.e. the hv_vmbus driver's init function returns -ENODEV due to vmbus_root_device being NULL).
In such a Linux VM, init_hv_pci_drv() runs with a side effect: the 3 hvpci_block_ops callbacks are set to functions that depend on hv_vmbus.
Later, when the MLX driver in such a VM invokes the callbacks, e.g. in drivers/net/ethernet/mellanox/mlx5/core/lib/hv.c:mlx5_hv_register_invalidate(), hvpci_block_ops.reg_blk_invalidate() is hv_register_block_invalidate() rather than a NULL function pointer, and hv_register_block_invalidate() assumes that it can find a struct hv_pcibus_device from pdev->bus->sysdata, which is false in such a VM.
Consequently, hv_register_block_invalidate() -> get_pcichild_wslot() -> spin_lock_irqsave() may hang since it can be accessing an invalid spinlock pointer.
Fix the issue by exporting hv_vmbus_exists() and using it in pci-hyperv:
hv_root_partition() is true and hv_nested is false ==>
hv_vmbus_exists() is false.
hv_root_partition() is true and hv_nested is true ==>
hv_vmbus_exists() is true.
hv_root_partition() is false ==> hv_vmbus_exists() is true.While at it, rename vmbus_exists() to hv_vmbus_exists() to follow the convention that all public functions have the hv_ prefix; also change the return value's type from int to bool to make the code more readable; also move the two pr_info() calls.
--- drivers/hv/vmbus_drv.c | 20 ++++++++------------ drivers/pci/controller/pci-hyperv.c | 2 +- include/linux/hyperv.h | 2 ++ 3 files changed, 11 insertions(+), 13 deletions(-)
drivers/hv/vmbus_drv.c +8 -12
@@ -101,13 +101,11 @@ struct device *hv_get_vmbus_root_device(void) }EXPORT_SYMBOL_GPL (hv_get_vmbus_root_device); -static int vmbus_exists (void ) +bool hv_vmbus_exists (void ) { -if (vmbus_root_device ==NULL ) -return -ENODEV ; - -return 0 ; +return vmbus_root_device !=NULL ; } +EXPORT_SYMBOL_GPL (hv_vmbus_exists);static u8 channel_monitor_group (const struct vmbus_channel *channel) {
@@ -1582,11 +1580,10 @@ int __vmbus_driver_register(struct hv_driver *hv_driver, struct module *owner, c {int ret; -pr_info ("registering driver %s\n" , hv_driver->name); +if (!hv_vmbus_exists ()) +return -ENODEV ; - ret =vmbus_exists (); -if (ret <0 ) -return ret; +pr_info ("registering driver %s\n" , hv_driver->name); hv_driver->driver.name = hv_driver->name; hv_driver->driver.owner = owner;
@@ -1612,9 +1609,8 @@ EXPORT_SYMBOL_GPL(__vmbus_driver_register);*/ void vmbus_driver_unregister (struct hv_driver *hv_driver) { -pr_info ("unregistering driver %s\n" , hv_driver->name); - -if (!vmbus_exists ()) { +if (hv_vmbus_exists ()) { +pr_info ("unregistering driver %s\n" , hv_driver->name);driver_unregister (&hv_driver->driver);vmbus_free_dynids (hv_driver); }
drivers/pci/controller/pci-hyperv.c +1 -1
@@ -4166,7 +4166,7 @@ static int __init init_hv_pci_drv(void)if (!hv_is_hyperv_initialized ())return -ENODEV ; -if (hv_root_partition () && !hv_nested) +if (!hv_vmbus_exists ())return -ENODEV ; ret =hv_pci_irqchip_init ();
include/linux/hyperv.h +2 -0
@@ -1304,6 +1304,8 @@ static inline void *hv_get_drvdata(struct hv_device *dev)struct device *hv_get_vmbus_root_device (void ); +bool hv_vmbus_exists (void ); +struct hv_ring_buffer_debug_info {u32 current_interrupt_mask;u32 current_read_index;
Signature · 1 line
2.43.0
Quoted 32 lines · Dexuan Cui
With commit
f84b21da3624("PCI: hv: Don't load the driver for baremetal root partition"), the bare metal Linux root partition won't use the pci-hyperv driver, but when a Linux VM runs on the Linux root partition, pci-hyperv'smodule_initfunctioninit_hv_pci_drv()can still run, e.g. in the case ofCONFIG_PCI_HYPERV=y, even if the VMBus driver is not used in such a VM (i.e. thehv_vmbusdriver's init function returns-ENODEVdue tovmbus_root_devicebeing NULL).In such a Linux VM,
init_hv_pci_drv()runs with a side effect: the 3hvpci_block_opscallbacks are set to functions that depend onhv_vmbus.Later, when the MLX driver in such a VM invokes the callbacks, e.g. in
drivers/net/ethernet/mellanox/mlx5/core/lib/hv.c:mlx5_hv_register_invalidate(),hvpci_block_ops.reg_blk_invalidate()ishv_register_block_invalidate()rather than a NULL function pointer, andhv_register_block_invalidate()assumes that it can find a structhv_pcibus_devicefrompdev->bus->sysdata, which is false in such a VM.Consequently,
hv_register_block_invalidate()->get_pcichild_wslot()->spin_lock_irqsave()may hang since it can be accessing an invalid spinlock pointer.Fix the issue by exporting
hv_vmbus_exists()and using it in pci-hyperv:hv_root_partition() is true and hv_nested is false ==> hv_vmbus_exists() is false. hv_root_partition() is true and hv_nested is true ==> hv_vmbus_exists() is true. hv_root_partition() is false ==> hv_vmbus_exists() is true.
Quoted 7 lines
While at it, rename
vmbus_exists()tohv_vmbus_exists()to follow the convention that all public functions have thehv_prefix; also change the return value's type from int to bool to make the code more readable; also move the twopr_info()calls.Reported-by Mukesh Rathor <[email protected]>Signed-off-by Dexuan Cui <[email protected]>
Applied. Thanks.