sync before umount; cleanup

This commit is contained in:
Glenn Maynard
2005-07-02 05:37:33 +00:00
parent 82e0d89f65
commit 0a3d0759b6
@@ -245,7 +245,7 @@ bool MemoryCardDriverThreaded_Linux::DoOneUpdate( bool bMount, vector<UsbStorage
FILEMAN->Unmount( "dir", d.sOsMountDir, TEMP_MOUNT_POINT );
}
ExecuteCommand( "umount -l \"" + d.sOsMountDir + "\"" );
ExecuteCommand( "sync; umount -l \"" + d.sOsMountDir + "\"" );
LOG->Trace( "WriteTest: %s, Name: %s", d.m_State == UsbStorageDevice::STATE_ERROR? "failed":"succeeded", d.sName.c_str() );
}
@@ -459,7 +459,7 @@ void MemoryCardDriverThreaded_Linux::Unmount( UsbStorageDevice* pDevice )
* by new devices until those are closed. Without this, if something
* causes the device to not unmount here, we'll never unmount it; that
* causes a device name leak, eventually running us out of mountpoints. */
CString sCommand = "umount -l \"" + pDevice->sDevice + "\"";
CString sCommand = "sync; umount -l \"" + pDevice->sDevice + "\"";
ExecuteCommand( sCommand );
}
@@ -469,11 +469,8 @@ void MemoryCardDriverThreaded_Linux::Flush( UsbStorageDevice* pDevice )
return;
// "sync" will only flush all file systems at the same time. -Chris
// I don't think so. Also, sync() merely queues a flush; it doesn't guarantee
// that the flush is completed on return. However, we can mount the filesystem
// with the flag "-o sync", which forces synchronous access (but that's probably
// very slow.) -glenn
ExecuteCommand( "mount -o remount " + pDevice->sDevice );
// That's OK.
ExecuteCommand( "sync" );
}
/*