On Tue. 9 May 2023 at 16:12, Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > On 09.05.2023 10:28:13, Vincent MAILHOL wrote: > > > > And because hdr.len is initially 3, hdr.len becomes 5. Right? Shouldn't it be 8? > > > > > > It might be a little confusing, but I think it's fine. > > > hdr.len is given in units of longwords (4 bytes each)! Therefore we > > > have 12 bytes (the initial 3 longwords) for struct tx_msg before > > > tx_msg.data[]. > > > Than (8 + 3)/4=2 gives us 2 additional longwords for the 8 data bytes. > > > So that 3+2=5 (equal to 20 bytes) should be ok. > > I think the term longword is more commonly used in non-Unix operating > systems :) > > > OK. So you want to round up the length to the next sizeof(long) multiple, right? > > > > First, sizeof(long) being platform specific, you need to declare a > > macro to make your intent explicit. > > > > /* Size of a long int on esd devices */ > > #define USB_ESD_SIZEOF_LONG 4 > > > > Please test, but for what I understand, below line is an equivalent > > and a more readable way to achieve your goal: > > > > msg->hdr.len = DIV_ROUND_UP(cf->len, USB_ESD_SIZEOF_LONG); > > use "sizeof(u32)" Marc is right, forget about USB_ESD_SIZEOF_LONG. sizeof(u32) does the job better.
