future IET development
A message that is not in the archive (sent off-list or not yet received): 1218829367.3145.31.camel@blackadder
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 23 lines · Ross S. W. Walker
Quoted 6 lines · rae l
I care that if IETD has some plan to push the iscsitarget kernel module (iscsitgt.ko) into the kenrel?
I think the iscsitarget kernel module is stable enough to submit into the kernel, if it's inlined in the kernel, everything will go better.
IET kernel module will not be included in the base Linux tree.
Other iSCSI modules have been included in the tree, and that is OK. At least for myself I do not see any real advantage to inclusion in the tree. If it were, then one would only find it in the latest kernel versions which one wouldn't see available in production systems for a number of years. Our development group is also too small for us to keep on top on the kernel development as we would need to do if it were included. We are quite happy keeping IET as a third party utility driver and hope to add some functionality for distribution maintainers to easily build packages for IET inclusion (.spec and .deb files).
Quoted 5 lines
Going forward, Arne does a brilliant job of keeping informed of the latest kernel API changes which allows us to keep IET working on the latest kernels, but we don't shadow the kernel developments on a day-by-day basis, so we recommend it best to run IET on mature kernel versions.
Have you read the Documentation/ of the kernel?
Didn't you know the benefits of getting the kernel code into the main kernel tree?
Here is some reasons from Documentation/stable_api_nonsense.txt:
The very good side effects of having your driver in the main kernel tree are:
- The quality of the driver will rise as the maintenance costs (to the original developer) will decrease.
- Other developers will add features to your driver.
- Other people will find and fix bugs in your driver.
- Other people will find tuning opportunities in your driver.
- Other people will update the driver for you when external interface changes require it.
- The driver automatically gets shipped in all Linux distributions without having to ask the distros to add it.
First, the demand for one iscsi target code inclusion into the main kernel tree is ongoing,
We need an iscsi target that can run on as much as possible kernels, not only the mature kernel versions.
In all kinds of our developments, the iscsi target is one basic functionality we want, we want iscsi can always be buildable at least, inclusion into the kernel is the best approach.
Second, I hope there will be more people who are interested in iscsi-target can test and review IET code, and improve IET to better. Indeed, in sometimes I observed IET kernel module not very stable on the latest 2.6.26 and 27-rcX kernels, sometimes it's kernel oops after serveral stop-ietd-unload-ko-load-ko-start-ietd loops, but not always reproducible, I'm still tracing on it;
the read-iscsi-consuming-100%-CPU is also another problem we observed in our lab.
All in one word, I hope more people can lay focus on IET to improve its quality. If we can improved it to production stable earlier, why not?
Last but not least significant, if you think iscsi target development group is too small to push the code into main kernel tree,
I can help on this, in fact, I have already a latest keep-up-with-linus git tree with iet kernel code integrated.
If someone is interested on this, we will collate and publish it some later.
Quoted 2 lines
-Ross
Thanks.
Signature · 5 lines
Denis Cheng Linux Application Developer "One of my most productive days was throwing away 1000 lines of code." - Ken Thompson.
Am Montag, den 01.09.2008, 01:36 +0800 schrieb rae l:
Quoted 5 lines
Have you read the
Documentation/of the kernel?
Didn't you know the benefits of getting the kernel code into the main kernel tree?Here is some reasons from
Documentation/stable_api_nonsense.txt:
[...]
You can safely assume that we're aware of the benefits of having the code in the kernel, but it will not happen. It does not make much sense to merge an iSCSI target into the kernel when there are generic target frameworks around that can also support different transports such as FC, FCOE. So there's not the remotest chance of getting IET merged.
Currently there's already STGT code in the kernel. It's basically an interface for kernel drivers such as FC, however most processing and many drivers (iSCSI, iSER, FCOE, ...) are done in userspace. And there's SCST. It also provides several transports and storage backends. It's mostly implemented in kernel space. Recently there's some movement to get it into the kernel.
If you're interested in more details just search the linux-scsi and the respective project's list archives.
Quoted 4 lines
Indeed, in sometimes I observed IET kernel module not very stable on the latest 2.6.26 and 27-rcX kernels, sometimes it's kernel oops after serveral stop-ietd-unload-ko-load-ko-start-ietd loops, but not always reproducible, I'm still tracing on it;
Please post the oops.
Thanks,
Arne