http://permalink.gmane.org/gmane.os.illumos.devel hourly 1 1901-01-01T00:00+00:00 http://gmane.org/img/gmane-25t.png http://gmane.org http://permalink.gmane.org/gmane.os.illumos.devel/12226 <pre>Yes please. - Garrett On May 23, 2013, at 12:55 PM, Matthew Ahrens &lt;mahrens&lt; at &gt;delphix.com&gt; wrote: ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Garrett D'Amore 2013-05-24T00:47:06 http://permalink.gmane.org/gmane.os.illumos.devel/12225 <pre> Last year Joerg (cc) was working on Illumos/COMSTAR VAAI patches, and he posted a couple of webrevs here on developer&lt; at &gt;lists.illumos.org. Joerg: Is it possible for you to publish those patches again? It'd be helpful for others to be able to take a look at the initial implementation, and hopefully collaborate! Thanks, </pre> Pasi Kärkkäinen 2013-05-23T20:20:37 http://permalink.gmane.org/gmane.os.illumos.devel/12224 <pre> FYI, ZFS support in the BSD bootloader is pretty robust -- it even supports booting from RAID-Z. Let me know if you want to be put in touch with the BSD guys who wrote/maintain this. --matt ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Matthew Ahrens 2013-05-23T19:55:19 http://permalink.gmane.org/gmane.os.illumos.devel/12223 <pre> first of all I'm NOT trying to advocate GPLv3 here :) "Ubuntu Switches Back To GRUB2 For SecureBoot": http://www.phoronix.com/scan.php?page=news_item&amp;px=MTE5MDI "Canonical's position was that the GPLv3-licensed boot-loader might lead them to needing to disclose their private SecureBoot key, thus compromising the support." "However, after Canonical talked it over with the Free Software Foundation and their OEM partners (along with taking heat from many others for moving away from GRUB2), they changed their minds" More info on the matter: http://blog.canonical.com/2012/09/20/quetzal-is-taking-flight-update-on-ubuntu-secure-boot-plans/ "In subsequent discussion with the Free Software Foundation (FSF), who owns the copyright for Grub 2, the FSF has stated clearly that Grub 2 with Secure Boot does not pose a risk of key disclosure in such circumstances. We have also confirmed that view with our OEM partners, and have introduced variations to the Ubuntu Certification program and QA scripts for pre-installs </pre> Pasi Kärkkäinen 2013-05-23T19:31:50 http://permalink.gmane.org/gmane.os.illumos.devel/12222 <pre>I do not really know, what copyrights should I add. Could anyone help me with that? I've already fixed stylistic errors. Cheers, Damian 2013/5/21 Alexander Eremin &lt;alexander.r.eremin&lt; at &gt;gmail.com&gt; ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Damian Bogel 2013-05-23T19:16:04 http://permalink.gmane.org/gmane.os.illumos.devel/12221 <pre>Sorry about possible cross-posting (tried sending to ZFS list, not sure it went through). I am investigating a hang that we have encountered in ZFS on servers with large memory (&gt;= 128GB) and small swap (1GB). Our ZFS pool consists of 12 mirrors, 2x1TB disks each, in a SAS-connected JBOD. There is a handful of datasets on this pool with record sizes varying from 4K to 128K. The hang is 100% reproducible using vdbench with our test script. Here's the scenario: The system goes low on memory (ARC uses up about 90GB) and VM slab allocator is quite fragmented - no contiguous 128K slab is available. A thread doing arc_read() is requesting 128K chunk of memory via zio_data_buf_alloc() and eventually kmem_cache_alloc() with KM_PUSHPAGE and is therefore blocked in vmem_xalloc() waiting on a cv. It is important to note that this thread is holding an HDR_LOCK on one of the buffer headers. The ARC eviction thread is trying to evict another buffer and free up some memory. In the process it calls arc_do_user_evicts()</pre> Ilya Usvyatsky 2013-05-23T18:20:55 http://permalink.gmane.org/gmane.os.illumos.devel/12220 <pre> While I am an open source developer you will not hear any arguments against the opinion you expressed from me since they match my own. It would be useful if Illumos can boot with both GRUB and GRUB2. For dual-boot machines, or due to current trends, GRUB2 may be the only option for some users. The option to be free from GPL encumberment will help Illumos, *BSD, and Apple's OS X to survive and thrive in the face of the popularity of Linux. Bob </pre> Bob Friesenhahn 2013-05-23T16:37:53 http://permalink.gmane.org/gmane.os.illumos.devel/12219 <pre> On May 23, 2013, at 7:15 AM, Bob Friesenhahn &lt;bfriesen&lt; at &gt;simple.dallas.tx.us&gt; wrote: No, it requires our customers to be able to replace the bootloader we use. We deliver an ISO image, making use of grub on the ISO, but we do not give customers the ability to build a new image. Furthermore, its possible to use hash checksums to make sure this image has *not* been modified, but use of GPLv3 would prevent us from doing that since it would require disclosing any private keys used in this process. This is also a potential stopper for building closed appliances -- think Delphix -- where they don't give customers access to the boot loader. You can argue this is Tivoization, which is what GPLv3 is designed to prevent. Nonetheless, there are often compelling business reasons to want to control your boot model from the ground up (tpm is part of secure solutions like this). So, while the GPLv3 advocates say this is what they want, my answer to them is -- no -- that is not what I want for illumos. I want illu</pre> Garrett D'Amore 2013-05-23T15:38:54 http://permalink.gmane.org/gmane.os.illumos.devel/12218 <pre> In what way does GPLv3 cause problems for you (or some users)? It would require the ability for a recipient to be able to build their own GRUB from source code and be able to install it (GPLv2 only required offering the source code). What other implications are you worried about? Large institutional users can distribute GPLv3 encumbered software within the institution without invoking GPL copying terms. Recently the US goverment decided that they are an institution so use of GPLv3 code within the US goverment does not invoke GPLv3 restrictions. As I recall, zfs support was already integrated into GRUB2. I am not sure what might make use of it. Maybe Linux on ZFS uses it. Bob </pre> Bob Friesenhahn 2013-05-23T14:15:45 http://permalink.gmane.org/gmane.os.illumos.devel/12217 <pre> Should have been more explicit, I meant what Christopher said, that cv_timedwait returns the "remaining time". All the manpage says about "&gt;0" is this: For cv_wait_sig(), cv_timedwait() or cv_timedwait_sig() indicates that the condition was met and the function returned due to a call to cv_signal() or cv_broadcast(), or due to a prema- ture wakeup (see NOTES). Nothing about what the return value actually means. Now if it really is the amount of time left on the clock, then great, I'll update the manpage. Cheers, </pre> Sašo Kiselkov 2013-05-22T19:59:14 http://permalink.gmane.org/gmane.os.illumos.devel/12216 <pre> It is documented at least as far as I can tell: http://illumos.org/man/9f/cv_timedwait. See the RETURN VALUES section. Robert ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Robert Mustacchi 2013-05-22T15:22:30 http://permalink.gmane.org/gmane.os.illumos.devel/12215 <pre> It is pushing the deadline only if the deadline was exceeded while waiting AND stmf_io_deadman_enabled is B_FALSE, i.e. the user wants us to wait indefinitely. In that case we can't just go ahead and offline the LU, because I/O tasks could still be referencing it. Interesting, I didn't notice the behavior of cv_timedwait documented anywhere (looked at the manpage - it's not there). Are you certain this is how cv_timedwait behaves? Shouldn't we document it then? (Maybe I just overlooked it, but I can't seem to find it mentioned in the "RETURN VALUES", "NOTES" or any other section). Thanks for taking the time review! </pre> Sašo Kiselkov 2013-05-22T07:21:02 http://permalink.gmane.org/gmane.os.illumos.devel/12214 <pre>Sorry to comment this late in the game, but I noticed one thing that seems like a bug: should you really be updating 'deadline' in the else clause in stmf_wait_ilu_tasks_finish() on line 8036. Isn't that just continually pushing the deadline into the future? And while I'm commenting I had one minor nit (feel free to ignore it): I don't think you need to get 'now' again and compare it to 'deadline' on line 8028 you can use the return value of cv_timedwait() which will return -1 if it is returning because the timeout expired or the remaining time (&gt;0) otherwise. Otherwise, LGTM. Chris On Tue, May 21, 2013 at 12:06 PM, Sašo Kiselkov &lt;skiselkov.ml&lt; at &gt;gmail.com&gt;wrote: ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.</pre> Christopher Siden 2013-05-21T23:06:37 http://permalink.gmane.org/gmane.os.illumos.devel/12213 <pre> Well, finally, it worked but with *export DMAKE_MAX_JOBS=1* I think I'll have to have a closer look to power needs in this box! Thanks for your help. Marc mosquito-[4]-illumos-gate% time ./nightly.sh illumos.sh real 311m42.594s user 256m24.582s sys 67m22.841s mosquito-[5]-illumos-gate% Le 20/05/13 23:16, Jim Klimov a écrit : It is a Gigabyte GV-N610SL-1GI - Graphics card - GF GT 610 - 1 GB DDR3 - PCIe 2.0 x16 low profile - DVI, D-Sub, HDMI - fanless 2 times 2TB: should need less power than the old 80 GB that used to be in this box ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Marc Lobelle 2013-05-21T20:17:39 http://permalink.gmane.org/gmane.os.illumos.devel/12212 <pre>On 05/21/2013 12:40 AM, Floris van Essen ..:: House of Ancients Amstafs ::.. wrote: Hi Floris, Unfortunately it doesn't. The existing VAAI support is Nexenta's proprietary code and it's up to them to release it, should they choose to do so. This only fixes some bugs in ALUA support (which I would love to see documented somewhat, but I'm not using it myself, so I don't know much about it). Cheers, </pre> Sašo Kiselkov 2013-05-21T19:05:51 http://permalink.gmane.org/gmane.os.illumos.devel/12211 <pre> Thanks, three reviews should be enough. Will RTI soon. Cheers, </pre> Sašo Kiselkov 2013-05-21T19:06:34 http://permalink.gmane.org/gmane.os.illumos.devel/12210 <pre>HI Saso, Does this include any form of VAAI ? That would be so so great :-) Cheers and keep up the good work, Floris -----Oorspronkelijk bericht----- Van: Sašo Kiselkov [mailto:skiselkov.ml&lt; at &gt;gmail.com] Verzonden: maandag 20 mei 2013 23:26 Aan: &lt;developer&lt; at &gt;lists.illumos.org&gt; Onderwerp: [developer] Review request #3775: Upstreaming of ALUA use-after-free fixes from Nexenta I'd like to upstream some ALUA fixes from Nexenta, graciously provided by Dan McDonald. Originally I planned on folding this into the fix for #3621 (ZFS LU stuck in the offlining state), but seeing as Delphix already did a fix for that, I will be reworking my fix for #3621 based on their solution. Therefore, I'd like to post this separately. Webrev: http://cr.illumos.org/~webrev/skiselkov/3775/ Cheers, -- Saso ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/23995791-de1e6769</pre> Floris van Essen ..:: House of Ancients Amstafs ::.. 2013-05-20T22:40:24 http://permalink.gmane.org/gmane.os.illumos.devel/12209 <pre>Hi all, Could this small fix be reviewed: http://cr.illumos.org/~webrev/bcoddens/illumos-3664/ Built on oi151a7, rebooted in the new BE and installed lynx browser as a test. Removed it again. Everything works fine. Best regards, Bart Coddens ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Bart Coddens 2013-05-21T17:07:34 http://permalink.gmane.org/gmane.os.illumos.devel/12208 <pre>Hi Folks, Webrev is here: http://cr.illumos.org/~webrev/bcoddens/illumos-3702/ Marcel, Could you provide some pointers how to test this fix ? Best Regards, Bart Coddens ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Bart Coddens 2013-05-21T17:10:14 http://permalink.gmane.org/gmane.os.illumos.devel/12207 <pre> Looks good Saso; I had reviewed Jeff's fix internally at Delphix previously, and the introduction of a deadman here seems fine. Ship it. -Seb ------------------------------------------- illumos-developer Archives: https://www.listbox.com/member/archive/182179/=now RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-86d49504 Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&amp;id_secret=21175072-abdf7b7e Powered by Listbox: http://www.listbox.com </pre> Sebastien Roy 2013-05-21T18:39:50 http://permalink.gmane.org/gmane.os.illumos.devel/12206 <pre> As I said, no testing is needed. It would be just a waste of your time. It is used only in these debug messages: http://src.illumos.org/source/s?refs=ct_str&amp;project=illumos-gate Full nightly is more than enough for this bug. Thanks. </pre> Marcel Telka 2013-05-21T17:21:49 Search the mailing list at Gmane query http://search.gmane.org/?group=$group=gmane.os.illumos.devel