> On Oct 30, 2018, at 22:55, 刘 轩 <Mr.liuxuan@xxxxxxxxxxx> wrote: > > please kick the website : > https://github.com/ceph/ceph-client/commit/c8a96a31cb04c7664626ab6ada7f66c98c09efbd > 1058 lines > I want to ask you question as follows: > Is spin_lock(&session->s_cap_lock) protect not needed by list_del(&cap->session_caps)？ > Spin_lock is not needed here because ‘dispose' is local variable. No one else can touch items in it > 发件人: Yan, Zheng [mailto:zyan@xxxxxxxxxx] > 发送时间: 2018年10月30日 20:34 > 收件人: ? ? > 抄送: ceph-users > 主题: Re: is it right involving cap->session_caps without lock protection in the two functions ? > > > > > On Oct 30, 2018, at 18:10, ? ? <Mr.liuxuan@xxxxxxxxxxx> wrote: > > Hello: > Recently, we have encountered a kernel dead question, and the reason we analyses vmcore dmesg is that list_add_tail(&cap->session_caps) in > __ceph_remove_cap has wrong,since &cap->session_cap is NULL! > so we analyses codes with cap->session_caps operaions, > We found these two functions involving cap - > session_cap operations, but there is no lock protection. > (1) cleanup_cap_releases > (2) ceph_send_cap_releases > SO, we want to ask you,whether there is a cap > session_caps operation exception in multithreading without locking or is it right code to operate cap->session_caps in these two functions without lock protection? and why if right? > > > There are protected by s_cap_lock. Which version of kernel do you use? _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |