From: Gal Pressman <galpress@amazon.com>
To: Michal Kalderon <mkalderon@marvell.com>
Cc: Ariel Elior <aelior@marvell.com>, "jgg@ziepe.ca" <jgg@ziepe.ca>,
"dledford@redhat.com" <dledford@redhat.com>,
"bmt@zurich.ibm.com" <bmt@zurich.ibm.com>,
"sleybo@amazon.com" <sleybo@amazon.com>,
"leon@kernel.org" <leon@kernel.org>,
"linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>
Subject: Re: [PATCH v7 rdma-next 3/7] RDMA/efa: Use the common mmap_xa helpers
Date: Sun, 25 Aug 2019 13:45:25 +0300 [thread overview]
Message-ID: <942f87b9-53dd-88ff-9045-2f3de7cc719c@amazon.com> (raw)
In-Reply-To: <MN2PR18MB3182C05D896D4FCA5CC86019A1A60@MN2PR18MB3182.namprd18.prod.outlook.com>
On 25/08/2019 11:41, Michal Kalderon wrote:
>>> @@ -515,46 +408,55 @@ static int qp_mmap_entries_setup(struct efa_qp
>> *qp,
>>> struct efa_com_create_qp_params
>> *params,
>>> struct efa_ibv_create_qp_resp *resp) {
>>> + u64 address;
>>> + u64 length;
>>> +
>>> /*
>>> * Once an entry is inserted it might be mmapped, hence cannot be
>>> * cleaned up until dealloc_ucontext.
>>> */
>>> resp->sq_db_mmap_key =
>>
>> Not a big deal, but now it makes more sense to assign qp-
>>> sq_db_mmap_key and assign the response later on.
> ok
>>
>>> - mmap_entry_insert(dev, ucontext, qp,
>>> - dev->db_bar_addr + resp->sq_db_offset,
>>> - PAGE_SIZE, EFA_MMAP_IO_NC);
>>> - if (resp->sq_db_mmap_key == EFA_MMAP_INVALID)
>>> + rdma_user_mmap_entry_insert(&ucontext->ibucontext, qp,
>>> + dev->db_bar_addr +
>>> + resp->sq_db_offset,
>>> + PAGE_SIZE, EFA_MMAP_IO_NC);
>>> + if (resp->sq_db_mmap_key == RDMA_USER_MMAP_INVALID)
>>> return -ENOMEM;
>>> -
>>> + qp->sq_db_mmap_key = resp->sq_db_mmap_key;
>>> resp->sq_db_offset &= ~PAGE_MASK;
>>>
>>> + address = dev->mem_bar_addr + resp->llq_desc_offset;
>>> + length = PAGE_ALIGN(params->sq_ring_size_in_bytes +
>>> + (resp->llq_desc_offset & ~PAGE_MASK));
>>> resp->llq_desc_mmap_key =
>>> - mmap_entry_insert(dev, ucontext, qp,
>>> - dev->mem_bar_addr + resp-
>>> llq_desc_offset,
>>> - PAGE_ALIGN(params-
>>> sq_ring_size_in_bytes +
>>> - (resp->llq_desc_offset &
>> ~PAGE_MASK)),
>>> - EFA_MMAP_IO_WC);
>>> - if (resp->llq_desc_mmap_key == EFA_MMAP_INVALID)
>>> + rdma_user_mmap_entry_insert(&ucontext->ibucontext, qp,
>>> + address,
>>> + length,
>>> + EFA_MMAP_IO_WC);
>>> + if (resp->llq_desc_mmap_key == RDMA_USER_MMAP_INVALID)
>>> return -ENOMEM;
>>> -
>>> + qp->llq_desc_mmap_key = resp->llq_desc_mmap_key;
>>> resp->llq_desc_offset &= ~PAGE_MASK;
>>>
>>> if (qp->rq_size) {
>>> + address = dev->db_bar_addr + resp->rq_db_offset;
>>> resp->rq_db_mmap_key =
>>> - mmap_entry_insert(dev, ucontext, qp,
>>> - dev->db_bar_addr + resp-
>>> rq_db_offset,
>>> - PAGE_SIZE, EFA_MMAP_IO_NC);
>>> - if (resp->rq_db_mmap_key == EFA_MMAP_INVALID)
>>> + rdma_user_mmap_entry_insert(&ucontext-
>>> ibucontext, qp,
>>> + address, PAGE_SIZE,
>>> + EFA_MMAP_IO_NC);
>>> + if (resp->rq_db_mmap_key ==
>> RDMA_USER_MMAP_INVALID)
>>> return -ENOMEM;
>>> -
>>> + qp->rq_db_mmap_key = resp->rq_db_mmap_key;
>>> resp->rq_db_offset &= ~PAGE_MASK;
>>>
>>> + address = virt_to_phys(qp->rq_cpu_addr);
>>> resp->rq_mmap_key =
>>> - mmap_entry_insert(dev, ucontext, qp,
>>> - virt_to_phys(qp->rq_cpu_addr),
>>> - qp->rq_size,
>> EFA_MMAP_DMA_PAGE);
>>> - if (resp->rq_mmap_key == EFA_MMAP_INVALID)
>>> + rdma_user_mmap_entry_insert(&ucontext-
>>> ibucontext, qp,
>>> + address, qp->rq_size,
>>> + EFA_MMAP_DMA_PAGE);
>>> + if (resp->rq_mmap_key == RDMA_USER_MMAP_INVALID)
>>> return -ENOMEM;
>>> + qp->rq_mmap_key = resp->rq_mmap_key;
>>>
>>> resp->rq_mmap_size = qp->rq_size;
>>> }
>>> @@ -775,6 +677,9 @@ struct ib_qp *efa_create_qp(struct ib_pd *ibpd,
>>> DMA_TO_DEVICE);
>>> if (!rq_entry_inserted)
>>
>> Now that we store the keys on the QP object we can remove the
>> rq_entry_inserted variable and test for !qp->rq_mmap_key.
> ok
>>
>>> free_pages_exact(qp->rq_cpu_addr, qp->rq_size);
>>> + else
>>> + rdma_user_mmap_entry_remove(&ucontext-
>>> ibucontext,
>>> + qp->rq_mmap_key);
>>
>> Other entries need to be removed as well, otherwise the refcount won't
>> reach zero. This error flow should now be similar to efa_destroy_qp. I think
>> that means losing the free_pages_exact too.
> Not sure I understand, how can we loose the free_pages_exact ? if the entry wasn’t
> Inserted into the mmap_xa what flow will free the pages ?
You're right.
Still need to remove other entries though.
next prev parent reply other threads:[~2019-08-25 10:45 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-20 12:18 [PATCH v7 rdma-next 0/7] RDMA/qedr: Use the doorbell overflow recovery mechanism for RDMA Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 1/7] RDMA/core: Move core content from ib_uverbs to ib_core Michal Kalderon
2019-08-20 12:58 ` Jason Gunthorpe
2019-08-20 21:30 ` Michal Kalderon
2019-08-21 16:30 ` Michal Kalderon
2019-08-20 14:08 ` Gal Pressman
2019-08-20 21:32 ` Michal Kalderon
2019-08-21 6:06 ` Gal Pressman
2019-08-21 7:56 ` Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 2/7] RDMA/core: Create mmap database and cookie helper functions Michal Kalderon
2019-08-20 13:21 ` Jason Gunthorpe
2019-08-20 21:23 ` [EXT] " Michal Kalderon
2019-08-21 16:47 ` Michal Kalderon
2019-08-21 16:51 ` Jason Gunthorpe
2019-08-21 17:14 ` Michal Kalderon
2019-08-21 17:37 ` Jason Gunthorpe
2019-08-26 11:53 ` Michal Kalderon
2019-08-26 12:01 ` Gal Pressman
2019-08-22 8:35 ` Gal Pressman
2019-08-25 8:36 ` Michal Kalderon
2019-08-25 10:39 ` Gal Pressman
2019-08-26 8:41 ` Michal Kalderon
2019-08-26 15:30 ` Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 3/7] RDMA/efa: Use the common mmap_xa helpers Michal Kalderon
2019-08-22 13:18 ` Gal Pressman
2019-08-25 8:41 ` Michal Kalderon
2019-08-25 10:45 ` Gal Pressman [this message]
2019-08-26 8:42 ` Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 4/7] RDMA/siw: " Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 5/7] RDMA/qedr: Use the common mmap API Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 6/7] RDMA/qedr: Add doorbell overflow recovery support Michal Kalderon
2019-08-20 12:18 ` [PATCH v7 rdma-next 7/7] RDMA/qedr: Add iWARP doorbell " Michal Kalderon
2019-08-20 18:31 ` [PATCH v7 rdma-next 0/7] RDMA/qedr: Use the doorbell overflow recovery mechanism for RDMA Gal Pressman
2019-08-21 8:03 ` Michal Kalderon
2019-08-21 10:15 ` Gal Pressman
2019-08-21 10:32 ` Michal Kalderon
2019-08-21 10:41 ` Gal Pressman
2019-08-21 12:25 ` Gal Pressman
2019-08-21 16:23 ` Gal Pressman
2019-08-21 16:27 ` Michal Kalderon
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=942f87b9-53dd-88ff-9045-2f3de7cc719c@amazon.com \
--to=galpress@amazon.com \
--cc=aelior@marvell.com \
--cc=bmt@zurich.ibm.com \
--cc=dledford@redhat.com \
--cc=jgg@ziepe.ca \
--cc=leon@kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=mkalderon@marvell.com \
--cc=sleybo@amazon.com \
/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).