On Sun, 01 Feb 2015, Michael Schmitz wrote: > Finn, Nicholas, > >> On Sat, 31 Jan 2015, Nicholas Mc Guire wrote: >> >>> This is only an API consolidation to make things more readable. >>> >>> Instances of var * HZ / 1000 are replaced by msecs_to_jiffies(var). >> >> ... and some instances of "value" are replaced by >> "msecs_to_jiffies(value)" >> which seems to be completely wrong. > > The values for USLEEP_* are taken to be in units jiffies, according to > comments in NCR5380.c. Replacing them by the msecs_to_jiffies conversion > is in fact wrong. > > Please drop the changes to g_NCR5380.c for that reason. > right the comment indicates it should be jiffies - my interpretation of that was that in NCR5380.c were jiffies + (250 * HZ / 1000); constructs would be correctly converted to e.g. jiffies + msecs_to_jiffies(250) And defines like USLEEP_POLL are noted to be in jiffies: * USLEEP_SLEEP - amount of time, in jiffies, to sleep and then defined correctly as HZ indepenedent values: #define USLEEP_SLEEP (20*HZ/1000) and thus should be the same as msecs_to_jiffies(20) now g_NCR5380.c defines #define USLEEP_POLL 1 #define USLEEP_SLEEP 20 #define USLEEP_WAITLONG 500 for the DTC3181E card - but without the conversion to ms from the use in the code though (e.g NCR5380_set_timer) it seemed to me that it actually should be jiffeis equivalent ms and the intent was only to change the USLEEP_POLL and USLEEP_WAITLONG settings for the specific device but not the unit and thus it shuld have been converted by msecs_to_jiffies(1) resp. msecs_to_jiffies(500). The problem with this if it is left in its current form is that the timeouts would actually depend on the HZ setting of the system which is probably not the intent. Pleas do give this one more look if the argument above is not sound I appologize for the noise. thx! hofrat -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
