From: Sagi Grimberg <sagig-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
To: Bart Van Assche
<bart.vanassche-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>,
Jason Gunthorpe
<jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
Cc: "linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org"
<linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
Subject: Re: [PATCH 7/8] IB/srp: Register the indirect data buffer descriptor
Date: Sun, 16 Aug 2015 14:09:10 +0300 [thread overview]
Message-ID: <55D06F56.4060005@dev.mellanox.co.il> (raw)
In-Reply-To: <55C93D21.1090102-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
On 8/11/2015 3:09 AM, Bart Van Assche wrote:
> Instead of always using the global rkey for the indirect data
> buffer descriptor, register that descriptor with the HCA if
> the kernel module parameter register_always has been set to Y.
>
> Signed-off-by: Bart Van Assche <bart.vanassche-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
> ---
> drivers/infiniband/ulp/srp/ib_srp.c | 57 +++++++++++++++++++++++++++++++++++--
> drivers/infiniband/ulp/srp/ib_srp.h | 4 +++
> 2 files changed, 58 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
> index 1029fd2..2a16ab4 100644
> --- a/drivers/infiniband/ulp/srp/ib_srp.c
> +++ b/drivers/infiniband/ulp/srp/ib_srp.c
> @@ -1453,18 +1453,58 @@ out:
> return ret;
> }
>
> +/*
> + * Register the indirect data buffer descriptor with the HCA.
> + *
> + * Note: since the indirect data buffer descriptor has been allocated with
> + * kmalloc() it is guaranteed that this buffer is a physically contiguous
> + * memory buffer.
> + */
kdoc style?
> +static int srp_map_idb(struct srp_rdma_ch *ch, struct srp_request *req,
> + void **next_mr, void **end_mr, u32 idb_len,
> + __be32 *idb_rkey)
> +{
> + struct srp_target_port *target = ch->target;
> + struct srp_device *dev = target->srp_host->srp_dev;
> + struct srp_map_state state;
> + struct srp_direct_buf idb_desc;
> + u64 idb_pages[1];
> + int ret;
> +
> + memset(&state, 0, sizeof(state));
> + memset(&idb_desc, 0, sizeof(idb_desc));
> + state.gen.next = next_mr;
> + state.gen.end = end_mr;
> + state.desc = &idb_desc;
> + state.pages = idb_pages;
> + state.pages[0] = (req->indirect_dma_addr &
> + dev->mr_page_mask);
> + state.npages = 1;
> + state.base_dma_addr = req->indirect_dma_addr;
> + state.dma_len = idb_len;
> + ret = srp_finish_mapping(&state, ch);
> + if (ret < 0)
> + goto out;
> +
> + *idb_rkey = idb_desc.key;
> +
> +out:
> + return ret;
> +}
> +
> static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
> struct srp_request *req)
> {
> struct srp_target_port *target = ch->target;
> struct scatterlist *scat;
> struct srp_cmd *cmd = req->cmd->buf;
> - int len, nents, count;
> + int len, nents, count, ret;
> struct srp_device *dev;
> struct ib_device *ibdev;
> struct srp_map_state state;
> struct srp_indirect_buf *indirect_hdr;
> - u32 table_len;
> + u32 idb_len, table_len;
> + __be32 idb_rkey;
> u8 fmt;
>
> if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
> @@ -1546,6 +1586,7 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
>
> count = min(state.ndesc, target->cmd_sg_cnt);
> table_len = state.ndesc * sizeof (struct srp_direct_buf);
> + idb_len = sizeof(struct srp_indirect_buf) + table_len;
>
> fmt = SRP_DATA_DESC_INDIRECT;
> len = sizeof(struct srp_cmd) + sizeof (struct srp_indirect_buf);
> @@ -1554,8 +1595,18 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_rdma_ch *ch,
> memcpy(indirect_hdr->desc_list, req->indirect_desc,
> count * sizeof (struct srp_direct_buf));
>
> + if (register_always && (dev->use_fast_reg || dev->use_fmr)) {
> + ret = srp_map_idb(ch, req, state.gen.next, state.gen.end,
> + idb_len, &idb_rkey);
> + if (ret < 0)
> + return ret;
> + req->nmdesc++;
> + } else {
> + idb_rkey = target->rkey;
> + }
> +
> indirect_hdr->table_desc.va = cpu_to_be64(req->indirect_dma_addr);
> - indirect_hdr->table_desc.key = cpu_to_be32(target->rkey);
> + indirect_hdr->table_desc.key = idb_rkey;
> indirect_hdr->table_desc.len = cpu_to_be32(table_len);
> indirect_hdr->len = cpu_to_be32(state.total_len);
>
> diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h
> index 60a33c1..255b0e5 100644
> --- a/drivers/infiniband/ulp/srp/ib_srp.h
> +++ b/drivers/infiniband/ulp/srp/ib_srp.h
> @@ -288,6 +288,10 @@ struct srp_map_state {
> struct srp_fr_desc **next;
> struct srp_fr_desc **end;
> } fr;
> + struct {
> + void **next;
> + void **end;
> + } gen;
What does 'gen' stands for?
--
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
next prev parent reply other threads:[~2015-08-16 11:09 UTC|newest]
Thread overview: 60+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-07-30 23:22 [PATCH v2 00/12] IB: Replace safe uses for ib_get_dma_mr with pd->local_dma_lkey Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 02/12] IB/mad: Remove ib_get_dma_mr calls Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 04/12] IB/mlx4: " Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 05/12] IB/mlx5: " Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 06/12] IB/iser: Use pd->local_dma_lkey Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 07/12] iser-target: Remove ib_get_dma_mr calls Jason Gunthorpe
[not found] ` <1438298547-21404-1-git-send-email-jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-07-30 23:22 ` [PATCH v2 01/12] IB/core: Guarantee that a local_dma_lkey is available Jason Gunthorpe
2015-08-02 13:09 ` Haggai Eran
2015-08-04 3:21 ` Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 03/12] IB/ipoib: Remove ib_get_dma_mr calls Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 08/12] IB/srp: Use pd->local_dma_lkey Jason Gunthorpe
2015-07-31 23:05 ` Bart Van Assche
2015-07-30 23:22 ` [PATCH v2 12/12] rds/ib: Remove ib_get_dma_mr calls Jason Gunthorpe
2015-08-14 2:47 ` santosh shilimkar
2015-07-30 23:22 ` [PATCH v2 09/12] IB/srp: Do not create an all physical insecure rkey by default Jason Gunthorpe
2015-08-03 15:39 ` Christoph Hellwig
2015-08-03 17:18 ` Bart Van Assche
2015-07-30 23:22 ` [PATCH v2 10/12] ib_srpt: Remove ib_get_dma_mr calls Jason Gunthorpe
2015-07-30 23:22 ` [PATCH v2 11/12] net/9p: " Jason Gunthorpe
2015-07-31 7:42 ` [PATCH v2 00/12] IB: Replace safe uses for ib_get_dma_mr with pd->local_dma_lkey Christoph Hellwig
2015-07-31 13:26 ` Steve Wise
2015-07-31 22:20 ` Bart Van Assche
[not found] ` <55BBF4B8.2050700-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-07-31 22:31 ` Jason Gunthorpe
[not found] ` <20150731223153.GA1518-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-07-31 23:04 ` Bart Van Assche
[not found] ` <55BBFF03.7000505-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-07-31 23:14 ` Jason Gunthorpe
[not found] ` <20150731231430.GA1955-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-07-31 23:32 ` Bart Van Assche
2015-08-01 20:05 ` Doug Ledford
[not found] ` <55BD2689.3080602-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2015-08-11 20:50 ` Bart Van Assche
[not found] ` <55CA600B.1050706-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-14 13:36 ` Doug Ledford
[not found] ` <55CDEEFA.4010803-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2015-08-14 16:20 ` Bart Van Assche
[not found] ` <55CE1554.60001-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-15 2:08 ` Doug Ledford
2015-08-03 15:24 ` Christoph Hellwig
[not found] ` <20150803152420.GA24193-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
2015-08-03 18:33 ` Bart Van Assche
[not found] ` <55BFB40F.8000500-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-04 18:09 ` Jason Gunthorpe
[not found] ` <20150804180933.GB5038-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-08-05 6:41 ` David Dillow
[not found] ` <1438756876.5698.2.camel-a7a0dvSY7KqLUyTwlgNVppKKF0rrzTr+@public.gmane.org>
2015-08-05 19:51 ` Jason Gunthorpe
[not found] ` <20150805195122.GA31595-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-08-05 21:45 ` Bart Van Assche
[not found] ` <55C2840C.5050301-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-05 22:41 ` Bart Van Assche
[not found] ` <55C2912A.50709-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-06 0:10 ` Jason Gunthorpe
[not found] ` <20150806001006.GD2483-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-08-06 0:19 ` Bart Van Assche
[not found] ` <55C2A7FE.7020904-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-06 4:36 ` Jason Gunthorpe
[not found] ` <20150806043642.GA14153-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-08-06 15:09 ` Bart Van Assche
2015-08-11 0:05 ` [PATCH 0/9] IB/srp: Do not create an all physical insecure rkey by default Bart Van Assche
[not found] ` <55C93C61.9010508-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-11 0:06 ` [PATCH 1/8] IB/srp: Re-enable FMR for non-page aligned buffers Bart Van Assche
[not found] ` <55C93C85.6090003-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-16 10:53 ` Sagi Grimberg
[not found] ` <55D06BB3.7070905-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2015-08-16 15:49 ` Bart Van Assche
2015-08-11 0:06 ` [PATCH 2/8] IB/srp: Use multiple registrations for large memory regions Bart Van Assche
2015-08-11 0:07 ` [PATCH 3/8] IB/srp: Add memory descriptor array pointer range checking Bart Van Assche
[not found] ` <55C93CBF.1060606-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-16 10:57 ` Sagi Grimberg
[not found] ` <55D06C9D.7030608-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2015-08-16 15:51 ` Bart Van Assche
2015-08-11 0:07 ` [PATCH 4/8] IB/srp: Remove the memory registration backtracking code Bart Van Assche
2015-08-11 0:08 ` [PATCH 5/8] IB/srp: Remove use_mr argument from srp_map_sg_entry() Bart Van Assche
2015-08-11 0:08 ` [PATCH 6/8] IB/srp: Introduce srp_device.use_fmr Bart Van Assche
[not found] ` <55C93D0C.7060000-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-16 11:03 ` Sagi Grimberg
[not found] ` <55D06E05.5060209-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2015-08-16 15:57 ` Bart Van Assche
2015-08-11 0:09 ` [PATCH 7/8] IB/srp: Register the indirect data buffer descriptor Bart Van Assche
[not found] ` <55C93D21.1090102-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org>
2015-08-16 11:09 ` Sagi Grimberg [this message]
[not found] ` <55D06F56.4060005-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2015-08-16 16:56 ` Bart Van Assche
2015-08-11 0:09 ` [PATCH 8/8] IB/srp: Create an insecure all physical rkey only if needed Bart Van Assche
2015-08-11 5:40 ` [PATCH 0/9] IB/srp: Do not create an all physical insecure rkey by default Jason Gunthorpe
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=55D06F56.4060005@dev.mellanox.co.il \
--to=sagig-ldsdmyg8hgv8yrgs2mwiifqbs+8scbdb@public.gmane.org \
--cc=bart.vanassche-XdAiOPVOjttBDgjK7y7TUQ@public.gmane.org \
--cc=jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).