MdeMdeModulePkg/UsbBusDxe: Fixed a possible memory leak bug introduced at r14226

The r14226 check-in indeed has memory leak in allocated "Child" pointer. UsbBusDriverBindingStop() may dereference this pointer and may bring exception on invalid memory access

Signed-off-by: Feng Tian <feng.tian@intel.com>
Reviewed-by: Jiewen Yao <jiewen.yao@intel.com>



git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@14251 6f19259b-4bc3-4df7-8a09-765794883524
This commit is contained in:
erictian 2013-04-07 08:43:28 +00:00
parent e79095b7b7
commit 7a4d52add1
1 changed files with 14 additions and 5 deletions

View File

@ -814,11 +814,20 @@ UsbEnumerateNewDev (
return EFI_SUCCESS; return EFI_SUCCESS;
ON_ERROR: ON_ERROR:
//
if (Child != NULL) { // If reach here, it means the enumeration process on a given port is interrupted due to error.
UsbFreeDevice (Child); // The s/w resources, including the assigned address(Address) and the allocated usb device data
} // structure(Bus->Devices[Address]), will NOT be freed here. These resources will be freed when
// the device is unplugged from the port or DriverBindingStop() is invoked.
//
// This way is used to co-work with the lower layer EDKII UHCI/EHCI/XHCI host controller driver.
// It's mainly because to keep UEFI spec unchanged EDKII XHCI driver have to maintain a state machine
// to keep track of the mapping between actual address and request address. If the request address
// (Address) is freed here, the Address value will be used by next enumerated device. Then EDKII XHCI
// host controller driver will have wrong information, which will cause further transaction error.
//
// EDKII UHCI/EHCI doesn't get impacted as it's make sense to reserve s/w resource till it gets unplugged.
//
return Status; return Status;
} }