Re: Use iscsi kernel passthrough interface for sendtargets
On Sun, 2009-08-30 at 14:31 -0700, Mike Christie wrote: For offlaod cards, bnx2i, cxgb3i and be2iscsi, we have to do discovery through a second network interface. The two attached patches has us instead do discovery through the iscsi class interface, so you do not have to set up a seperate nic for discovery and you do not have to worry about IP based ACLS on the target. The 0001-iscsiadm-use-kernel-pdu-passthrough-for-sendtargets.patch patch was made over the open-iscsi git tree. It is only userspce code. The 0001-libiscsi-fix-login-text-checks-in-pdu-injection-cod.patch patch is made over linux-2.6-iscsi's iscsi branch. It should probably apply to other kernels though. Please give the patches a try. Here are some known limits: MaxRecvSegmentLength is hard coded to 8K for discovery sessions, because the kernel conn-data buffer and driver related buffers were hard coded to this value and some netlink's buffers are too. I have only tested cxgb3i. Ben, this is not patched against the uip offload tree but I can do that for you. Michael and Anil, I was not sure where to handle text response pdus in bnx2i_process_new_cqes? I am currently sending the discovery session login and text pdus through the same interface you get nops or normal session login pdus, so I think bnx2i_mtask_xmit should just work (I limited MaxRecvSegmentLength to ISCSI_DEF_MAX_RECV_SEG_LEN, so the buf sizes are ok.). I will work on it and send it to you soon. where is the latest bnx2i code on your linux-2.6-iscsi tree? Jay, not sure how text response pdus are coming in? Are they like login or logout pdus? --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---
Re: Use iscsi kernel passthrough interface for sendtargets
On 09/01/2009 01:14 PM, Anil Veerabhadrappa wrote: On Sun, 2009-08-30 at 14:31 -0700, Mike Christie wrote: For offlaod cards, bnx2i, cxgb3i and be2iscsi, we have to do discovery through a second network interface. The two attached patches has us instead do discovery through the iscsi class interface, so you do not have to set up a seperate nic for discovery and you do not have to worry about IP based ACLS on the target. The 0001-iscsiadm-use-kernel-pdu-passthrough-for-sendtargets.patch patch was made over the open-iscsi git tree. It is only userspce code. The 0001-libiscsi-fix-login-text-checks-in-pdu-injection-cod.patch patch is made over linux-2.6-iscsi's iscsi branch. It should probably apply to other kernels though. Please give the patches a try. Here are some known limits: MaxRecvSegmentLength is hard coded to 8K for discovery sessions, because the kernel conn-data buffer and driver related buffers were hard coded to this value and some netlink's buffers are too. I have only tested cxgb3i. Ben, this is not patched against the uip offload tree but I can do that for you. Michael and Anil, I was not sure where to handle text response pdus in bnx2i_process_new_cqes? I am currently sending the discovery session login and text pdus through the same interface you get nops or normal session login pdus, so I think bnx2i_mtask_xmit should just work (I limited MaxRecvSegmentLength to ISCSI_DEF_MAX_RECV_SEG_LEN, so the buf sizes are ok.). I will work on it and send it to you soon. where is the latest bnx2i code on your linux-2.6-iscsi tree? The latest bnx2i code is in James bnx2i tree. linux-2.6-iscsi does not have your cid changes yet. I am updating to James's tree now. However, the kernel patch will apply to James's tree as is, so you can go ahead and use it. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---
Re: Use iscsi kernel passthrough interface for sendtargets
On 09/01/2009 01:28 PM, Mike Christie wrote: The latest bnx2i code is in James bnx2i tree. linux-2.6-iscsi does not have your cid changes yet. I am updating to James's tree now. However, the kernel patch will apply to James's tree as is, so you can go ahead and use it. It looks like James's tree is missing the GFP_ATOMIC iscsi_offload_msg patch you guys did because that went through Dave. It might be missing other bnx2 stuff too. How are you guys syncing your stuff when it some of it is in Dave's tree and some of it is in James's tree? Do you just figure out where it is then pull from each tree? --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---
Re: Use iscsi kernel passthrough interface for sendtargets
On 09/01/2009 01:37 PM, Mike Christie wrote: On 09/01/2009 01:28 PM, Mike Christie wrote: The latest bnx2i code is in James bnx2i tree. linux-2.6-iscsi does not have your cid changes yet. I am updating to James's tree now. However, the kernel patch will apply to James's tree as is, so you can go ahead and use it. It looks like James's tree is missing the GFP_ATOMIC iscsi_offload_msg patch you guys did because that went through Dave. It might be missing other bnx2 stuff too. I updated the linux-2.6-iscsi branch to linus's tree and took in your bnx2i patches in James's scsi-misc tree. I did not grab the bnx2/cnic patches by accident. I just remembered Dave picked them up (I was just searching linus and scsi-misc/scsi-rc-fixes). In the iscsi branch I have outstanding iscsi patches and the kernel patch for this (I also included the suspend queue patch from yesterday). The last commit is: commit 35074fae1a1cf6eb1a22e912d16ac35ecd9020c8 Author: Mike Christie micha...@cs.wisc.edu Date: Sun Aug 30 16:19:44 2009 -0500 libiscsi: fix login/text checks in pdu injection code --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---
Re: Use iscsi kernel passthrough interface for sendtargets
On 08/30/2009 09:57 PM, Benjamin Li wrote: Hi Mike, I am still looking over/understanding the patches but could I trouble you patch this against the uIP offload tree? Will do. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---
Re: Use iscsi kernel passthrough interface for sendtargets
Hi Mike, I am still looking over/understanding the patches but could I trouble you patch this against the uIP offload tree? Thanks again. -Ben On Sun, 2009-08-30 at 14:31 -0700, Mike Christie wrote: For offlaod cards, bnx2i, cxgb3i and be2iscsi, we have to do discovery through a second network interface. The two attached patches has us instead do discovery through the iscsi class interface, so you do not have to set up a seperate nic for discovery and you do not have to worry about IP based ACLS on the target. The 0001-iscsiadm-use-kernel-pdu-passthrough-for-sendtargets.patch patch was made over the open-iscsi git tree. It is only userspce code. The 0001-libiscsi-fix-login-text-checks-in-pdu-injection-cod.patch patch is made over linux-2.6-iscsi's iscsi branch. It should probably apply to other kernels though. Please give the patches a try. Here are some known limits: MaxRecvSegmentLength is hard coded to 8K for discovery sessions, because the kernel conn-data buffer and driver related buffers were hard coded to this value and some netlink's buffers are too. I have only tested cxgb3i. Ben, this is not patched against the uip offload tree but I can do that for you. Michael and Anil, I was not sure where to handle text response pdus in bnx2i_process_new_cqes? I am currently sending the discovery session login and text pdus through the same interface you get nops or normal session login pdus, so I think bnx2i_mtask_xmit should just work (I limited MaxRecvSegmentLength to ISCSI_DEF_MAX_RECV_SEG_LEN, so the buf sizes are ok.). Jay, not sure how text response pdus are coming in? Are they like login or logout pdus? --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups open-iscsi group. To post to this group, send email to open-iscsi@googlegroups.com To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/open-iscsi -~--~~~~--~~--~--~---