Hi, Jeremie and Lists. The previous patch to change if_em's int_throttle_ceil into sysctl-able cause a kernel panic. If you set em's int_throttle_ceil=0 and then reconfigure the em, it cause a diveded by zero panic. This patch for original if_em.[ch] which is attached to this mail is corrected this problem. -- Shunsuke SHINOMIYA <shino@fornext.org> -------------- next part -------------- A non-text attachment was scrubbed... Name: if_em.diff Type: application/octet-stream Size: 3458 bytes Desc: not available Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20041210/cd920769/if_em.obj
On Fri, Dec 10, 2004 at 12:00:21PM +0900, Shunsuke SHINOMIYA wrote:> The previous patch to change if_em's int_throttle_ceil into > sysctl-able cause a kernel panic. If you set em's int_throttle_ceil=0 > and then reconfigure the em, it cause a diveded by zero panic. > > This patch for original if_em.[ch] which is attached to this mail is > corrected this problem.Great work Shunsuke, thank you ! Any chance to get this patch reviewed and commited ? Although Bjoern Zeeb proposed to use a global way to control interrupt moderation among network drivers, there was no proposition to unify it. It would be a pity if this patch gets lost in the archives without being either commited or submited as a PR. Regards, -- Jeremie Le Hen jeremie@le-hen.org
Hi Jeremie and Lists,> Great work Shunsuke, thank you ! > Any chance to get this patch reviewed and commited ?Thank you. I did send-pr as kern/74986. http://www.freebsd.org/cgi/query-pr.cgi?pr=74986 I think that the appropriate value of interrupt moderation depends on each system, but tuning this value is worth to bring out a performance of the system. I hope that this patch help to bring out the performance of your system. -- Shunsuke SHINOMIYA <shino@fornext.org>