From: Sagi Grimberg <sagig-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> To: "Hefty, Sean" <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>, Chuck Lever <chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> Cc: "linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" <linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>, Linux NFS Mailing List <linux-nfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org> Subject: Re: [PATCH v1 04/14] xprtrdma: Use ib_device pointer safely Date: Tue, 12 May 2015 13:01:55 +0300 [thread overview] Message-ID: <5551CF93.4070501@dev.mellanox.co.il> (raw) In-Reply-To: <1828884A29C6694DAF28B7E6B8A82373A8FCE01F-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org> On 5/11/2015 9:26 PM, Hefty, Sean wrote: >>> ia->ri_device is never updated. The only place it is set is in >>> rpcrdma_ia_open(). >> >> So you assume that each ri_id that you will recreate contains the >> same device handle? >> >> I think that for ADDR_CHANGE event when the slave belongs to another >> device you will hit a mismatch. CC'ing Sean for more info... > > I'm not familiar with the xprtrdma code. From the perspective of the rdma_cm, if a listen is associated with a specific IP address, then it will also be associated with a specific device. If an address change occurs, and the address moves to another device, then the app is essentially left with an unusable listen. Received connection requests will not find a matching listen and be dropped. > > If the address moves ports on the same device, then I think this works out fine in the case where the app ignores the ADDR_CHANGE event. > >>> It might be saner to construct a virtual ib_device in this >>> case that consumers can depend on. >> >> I'm not sure how does a virtual ib_device can work - that goes >> to the verbs themselves... Seems like a layering mis-match to me... > > I interpreted Chuck as asking for some lower-level way of handling this, so that all ULPs don't need to re-implement this. It's not really a re-implementation, address resolution attaches the the ib device. The assumption that for the same address the same ib device will be attached is wrong for bonding over more than one device. It's just something that rdma_cm user needs to be aware of if/when trying to cache the device handle. > I agree that this would be hard (impossible?) without ULPs operating at a slightly higher level of abstraction. > Maybe the rdma_cm could be extended for this purpose, for example, by exposing its own devices, which either map directly to a verbs device, or handle a virtual mapping. I guess it can be done, but it will require port all the ULPs to use some sort of abstraction layer. Something that I'm not very enthusiastic about... Sagi. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html
WARNING: multiple messages have this Message-ID (diff)
From: Sagi Grimberg <sagig@dev.mellanox.co.il> To: "Hefty, Sean" <sean.hefty@intel.com>, Chuck Lever <chuck.lever@oracle.com> Cc: "linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>, Linux NFS Mailing List <linux-nfs@vger.kernel.org> Subject: Re: [PATCH v1 04/14] xprtrdma: Use ib_device pointer safely Date: Tue, 12 May 2015 13:01:55 +0300 [thread overview] Message-ID: <5551CF93.4070501@dev.mellanox.co.il> (raw) In-Reply-To: <1828884A29C6694DAF28B7E6B8A82373A8FCE01F@ORSMSX109.amr.corp.intel.com> On 5/11/2015 9:26 PM, Hefty, Sean wrote: >>> ia->ri_device is never updated. The only place it is set is in >>> rpcrdma_ia_open(). >> >> So you assume that each ri_id that you will recreate contains the >> same device handle? >> >> I think that for ADDR_CHANGE event when the slave belongs to another >> device you will hit a mismatch. CC'ing Sean for more info... > > I'm not familiar with the xprtrdma code. From the perspective of the rdma_cm, if a listen is associated with a specific IP address, then it will also be associated with a specific device. If an address change occurs, and the address moves to another device, then the app is essentially left with an unusable listen. Received connection requests will not find a matching listen and be dropped. > > If the address moves ports on the same device, then I think this works out fine in the case where the app ignores the ADDR_CHANGE event. > >>> It might be saner to construct a virtual ib_device in this >>> case that consumers can depend on. >> >> I'm not sure how does a virtual ib_device can work - that goes >> to the verbs themselves... Seems like a layering mis-match to me... > > I interpreted Chuck as asking for some lower-level way of handling this, so that all ULPs don't need to re-implement this. It's not really a re-implementation, address resolution attaches the the ib device. The assumption that for the same address the same ib device will be attached is wrong for bonding over more than one device. It's just something that rdma_cm user needs to be aware of if/when trying to cache the device handle. > I agree that this would be hard (impossible?) without ULPs operating at a slightly higher level of abstraction. > Maybe the rdma_cm could be extended for this purpose, for example, by exposing its own devices, which either map directly to a verbs device, or handle a virtual mapping. I guess it can be done, but it will require port all the ULPs to use some sort of abstraction layer. Something that I'm not very enthusiastic about... Sagi.
next prev parent reply other threads:[~2015-05-12 10:01 UTC|newest] Thread overview: 114+ messages / expand[flat|nested] mbox.gz Atom feed top 2015-05-04 17:56 [PATCH v1 00/14] client NFS/RDMA patches for 4.2 Chuck Lever 2015-05-04 17:56 ` Chuck Lever [not found] ` <20150504174626.3483.97639.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-04 17:56 ` [PATCH v1 01/14] xprtrdma: Transport fault injection Chuck Lever 2015-05-04 17:56 ` Chuck Lever [not found] ` <20150504175651.3483.35554.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-05 13:49 ` Anna Schumaker 2015-05-05 13:49 ` Anna Schumaker [not found] ` <5548CA82.9060903-ZwjVKphTwtPQT0dZR+AlfA@public.gmane.org> 2015-05-05 13:53 ` Chuck Lever 2015-05-05 13:53 ` Chuck Lever [not found] ` <06C797AA-4E45-49DA-8CA8-2B8D22B818DB-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-05 14:44 ` Anna Schumaker 2015-05-05 14:44 ` Anna Schumaker [not found] ` <5548D741.7090803-ZwjVKphTwtPQT0dZR+AlfA@public.gmane.org> 2015-05-05 15:15 ` Chuck Lever 2015-05-05 15:15 ` Chuck Lever [not found] ` <2098B4A5-48C7-4458-BAC6-10F64359C405-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-05 15:16 ` Anna Schumaker 2015-05-05 15:16 ` Anna Schumaker 2015-05-05 15:10 ` Steve Wise 2015-05-05 15:10 ` Steve Wise 2015-05-04 17:57 ` [PATCH v1 02/14] xprtrdma: Warn when there are orphaned IB objects Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175700.3483.57728.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-06 11:37 ` Devesh Sharma 2015-05-06 11:37 ` Devesh Sharma [not found] ` <CANjDDBicLDb7K=RZB5Gqv+oACaAG8DANJX3ML971jnXd_5KEuQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org> 2015-05-06 13:24 ` Chuck Lever 2015-05-06 13:24 ` Chuck Lever [not found] ` <963F9850-38D0-4434-88E8-14BC42F74499-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-06 14:05 ` Sagi Grimberg 2015-05-06 14:05 ` Sagi Grimberg 2015-05-06 14:22 ` Devesh Sharma 2015-05-06 14:22 ` Devesh Sharma [not found] ` <CANjDDBg3Ey0DEM3n9PY9VLW+chF0=AJPghYOopo1a9vL3CKzKg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org> 2015-05-06 16:48 ` Jason Gunthorpe 2015-05-06 16:48 ` Jason Gunthorpe [not found] ` <20150506164817.GC11331-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org> 2015-05-07 7:53 ` Devesh Sharma 2015-05-07 7:53 ` Devesh Sharma 2015-05-04 17:57 ` [PATCH v1 03/14] xprtrdma: Replace rpcrdma_rep::rr_buffer with rr_rxprt Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175711.3483.17222.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 9:38 ` Sagi Grimberg 2015-05-07 9:38 ` Sagi Grimberg [not found] ` <554B328B.9040804-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-07 13:25 ` Chuck Lever 2015-05-07 13:25 ` Chuck Lever 2015-05-04 17:57 ` [PATCH v1 04/14] xprtrdma: Use ib_device pointer safely Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175720.3483.80356.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:00 ` Sagi Grimberg 2015-05-07 10:00 ` Sagi Grimberg [not found] ` <554B37CF.2070206-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-07 13:39 ` Chuck Lever 2015-05-07 13:39 ` Chuck Lever [not found] ` <E1ADA91B-45DA-46B6-A114-E2600613969A-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-07 13:56 ` Sagi Grimberg 2015-05-07 13:56 ` Sagi Grimberg [not found] ` <554B6F2A.6000608-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-07 14:12 ` Chuck Lever 2015-05-07 14:12 ` Chuck Lever [not found] ` <BDE22240-DC37-4C54-B71E-D88EF54D3119-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-07 15:11 ` Sagi Grimberg 2015-05-07 15:11 ` Sagi Grimberg [not found] ` <554B80B7.8090900-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-11 15:22 ` Chuck Lever 2015-05-11 15:22 ` Chuck Lever 2015-05-11 18:26 ` Hefty, Sean 2015-05-11 18:26 ` Hefty, Sean [not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FCE01F-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org> 2015-05-11 18:57 ` Chuck Lever 2015-05-11 18:57 ` Chuck Lever 2015-05-12 10:01 ` Sagi Grimberg [this message] 2015-05-12 10:01 ` Sagi Grimberg 2015-05-04 17:57 ` [PATCH v1 05/14] xprtrdma: Introduce helpers for allocating MWs Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175730.3483.51996.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:16 ` Sagi Grimberg 2015-05-07 10:16 ` Sagi Grimberg 2015-05-04 17:57 ` [PATCH v1 06/14] xprtrdma: Acquire FMRs in rpcrdma_fmr_register_external() Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175739.3483.46010.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:15 ` Sagi Grimberg 2015-05-07 10:15 ` Sagi Grimberg 2015-05-04 17:57 ` [PATCH v1 07/14] xprtrdma: Introduce an FRMR recovery workqueue Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175749.3483.56570.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:37 ` Devesh Sharma 2015-05-07 10:37 ` Devesh Sharma 2015-05-04 17:57 ` [PATCH v1 08/14] xprtrdma: Acquire MRs in rpcrdma_register_external() Chuck Lever 2015-05-04 17:57 ` Chuck Lever [not found] ` <20150504175758.3483.44890.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:31 ` Sagi Grimberg 2015-05-07 10:31 ` Sagi Grimberg [not found] ` <554B3EEB.7070302-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:24 ` Devesh Sharma 2015-05-08 15:24 ` Devesh Sharma [not found] ` <CANjDDBiGLcaAofGwz6OGEXUUE_b2rcZepv0ebvTc-XNVEBq5Mw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org> 2015-05-08 15:40 ` Chuck Lever 2015-05-08 15:40 ` Chuck Lever [not found] ` <6FBAAAF3-3E70-418F-A887-C022525D6C4F-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> 2015-05-10 10:17 ` Sagi Grimberg 2015-05-10 10:17 ` Sagi Grimberg 2015-05-04 17:58 ` [PATCH v1 09/14] xprtrdma: Remove unused LOCAL_INV recovery logic Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175808.3483.57643.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:35 ` Sagi Grimberg 2015-05-07 10:35 ` Sagi Grimberg [not found] ` <554B3FFA.5020101-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:31 ` Devesh Sharma 2015-05-08 15:31 ` Devesh Sharma 2015-05-04 17:58 ` [PATCH v1 10/14] xprtrdma: Remove ->ro_reset Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175818.3483.22408.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:36 ` Sagi Grimberg 2015-05-07 10:36 ` Sagi Grimberg [not found] ` <554B402F.3000604-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:33 ` Devesh Sharma 2015-05-08 15:33 ` Devesh Sharma 2015-05-04 17:58 ` [PATCH v1 11/14] xprtrdma: Remove rpcrdma_ia::ri_memreg_strategy Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175827.3483.62904.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:36 ` Sagi Grimberg 2015-05-07 10:36 ` Sagi Grimberg [not found] ` <554B404B.9090301-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:34 ` Devesh Sharma 2015-05-08 15:34 ` Devesh Sharma 2015-05-04 17:58 ` [PATCH v1 12/14] xprtrdma: Split rb_lock Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175837.3483.28838.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:37 ` Sagi Grimberg 2015-05-07 10:37 ` Sagi Grimberg 2015-05-04 17:58 ` [PATCH v1 13/14] xprtrdma: Stack relief in fmr_op_map() Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175846.3483.32959.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 10:50 ` Sagi Grimberg 2015-05-07 10:50 ` Sagi Grimberg [not found] ` <554B436B.5040108-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:36 ` Devesh Sharma 2015-05-08 15:36 ` Devesh Sharma 2015-05-04 17:58 ` [PATCH v1 14/14] xprtrmda: Reduce per-transport MR allocation Chuck Lever 2015-05-04 17:58 ` Chuck Lever [not found] ` <20150504175856.3483.57373.stgit-FYjufvaPoItvLzlybtyyYzGyq/o6K9yX@public.gmane.org> 2015-05-07 11:00 ` Sagi Grimberg 2015-05-07 11:00 ` Sagi Grimberg [not found] ` <554B45E7.4040902-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> 2015-05-08 15:53 ` Devesh Sharma 2015-05-08 15:53 ` Devesh Sharma 2015-05-05 15:17 ` [PATCH v1 00/14] client NFS/RDMA patches for 4.2 Steve Wise 2015-05-05 15:17 ` Steve Wise
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=5551CF93.4070501@dev.mellanox.co.il \ --to=sagig-ldsdmyg8hgv8yrgs2mwiifqbs+8scbdb@public.gmane.org \ --cc=chuck.lever-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org \ --cc=linux-nfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \ --cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \ --cc=sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.