On 06/12/2017 05:51 PM, Mike Christie wrote: > On 06/12/2017 03:03 PM, Bart Van Assche wrote: >> On Mon, 2017-06-12 at 13:40 -0500, Mike Christie wrote: >>> For the xcopy use, I was still trying to figure out how that worked >>> before my patches. target_xcopy_locate_se_dev_e4 would return a device >>> under the g_device_mutex but I could not figure out how it was protected >>> from removal after that mutex was dropped. >> >> Hello Mike, >> >> I don't think that protection did already exist. Had you noticed the >> following patch: "[PATCH v6 12/33] target: Introduce target_get_device() >> and target_put_device()" >> (https://www.spinics.net/lists/target-devel/msg14535.html)? >> > > Thanks. > > What about target_depend_item/target_undepend_item calls? That it would > prevent removals while the device is in use. > > You would need get/put calls if you used my lookup/iter helpers while > removal was already in progress and could race. > > Do we want to do both for my next posting? For the current uses (tcmu > and xcopy), both are not needed I do not think, but do we want to make > the interface for generic use? For the latter, I will build off of your > patches then. Oops. I actually need it in the removal path, so I cannot use target_depend_item, so I guess we need to use refcounts. I will do some testing and code review to see how xcopy works when we allow removals and only get a refcount to the device. -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
