Re: [PATCH 1/2] usbip: Fix seqnum sign extension issue in vhci_tx_urb

[Shuah Khan]

On 1/3/25 08:18, xndcn wrote:
> Thanks.
>
> > How did you find the problem?
> > Why does it make sense to cast it to u32?
>
> After running with usbip enough time, I happened to see logs like this:
> > [ 293.863125] vhci_hcd vhci_hcd.0: Device attached
> > [ 294.081110] usb 1-1: new high-speed USB device number 2 using vhci_hcd
> > [ 294.193163] usb 1-1: SetAddress Request (2) to port 0
> > [ 294.204334] vhci_hcd: cannot find a urb of seqnum 2147483648 max seqnum -2147483648
> > [ 294.204850] vhci_hcd: stop threads
> > [ 294.204851] vhci_hcd: release socket
> > [ 294.204853] vhci_hcd: disconnect device
>
> Then I notice that on 64bit platform, when
> atomic_inc_return(&vhci_hcd->seqnum) returns (2147483647 + 1, or
> 0x80000000),
> priv->seqnum (which is unsigned long, i.e. u64 on 64bit) will be
> extends to 0xffffffff80000000
> So we can fix the issue by cast it to u32.

Can you send me the dmesg without and with your patch?

thanks,
-- Shuah