From: "Saleem, Shiraz" <shiraz.saleem@intel.com>
To: Jason Gunthorpe <jgg@mellanox.com>, Gal Pressman <galpress@amazon.com>
Cc: Greg KH <gregkh@linuxfoundation.org>,
"Kirsher, Jeffrey T" <jeffrey.t.kirsher@intel.com>,
"dledford@redhat.com" <dledford@redhat.com>,
"davem@davemloft.net" <davem@davemloft.net>,
"Ismail, Mustafa" <mustafa.ismail@intel.com>,
"linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>,
"netdev@vger.kernel.org" <netdev@vger.kernel.org>,
"nhorman@redhat.com" <nhorman@redhat.com>,
"sassmann@redhat.com" <sassmann@redhat.com>,
"poswald@suse.com" <poswald@suse.com>
Subject: RE: [RDMA RFC v6 14/16] RDMA/irdma: Add ABI definitions
Date: Wed, 27 May 2020 01:58:14 +0000 [thread overview]
Message-ID: <9DD61F30A802C4429A01CA4200E302A7EE04048D@fmsmsx124.amr.corp.intel.com> (raw)
In-Reply-To: <20200520123726.GD24561@mellanox.com>
> Subject: Re: [RDMA RFC v6 14/16] RDMA/irdma: Add ABI definitions
>
> On Wed, May 20, 2020 at 12:02:35PM +0300, Gal Pressman wrote:
> > On 20/05/2020 11:52, Greg KH wrote:
> > > On Wed, May 20, 2020 at 10:54:25AM +0300, Gal Pressman wrote:
> > >> On 20/05/2020 10:04, Jeff Kirsher wrote:
> > >>> +struct i40iw_create_qp_resp {
> > >>> + __u32 qp_id;
> > >>> + __u32 actual_sq_size;
> > >>> + __u32 actual_rq_size;
> > >>> + __u32 i40iw_drv_opt;
> > >>> + __u16 push_idx;
> > >>> + __u8 lsmm;
> > >>> + __u8 rsvd;
> > >>> +};
> > >>
> > >> This struct size should be 8 bytes aligned.
> > >
> > > Aligned in what way? Seems sane to me, what would you want it to
> > > look like instead?
> >
> > The uverbs ABI structs sizes are assumed to be padded to 8 bytes
> > alignment, I would expect the reserved field to be an array of 5 bytes
> > as done in other structs in this file (irdma_modify_qp_req for example).
> > Jason could correct me if I'm wrong?
>
> "it is complicated"
>
> The udata structs must have alignment that is compatible with the core struct that
> prefixes them. Of course we have a mess here, and nothing is uniform..
>
> In this case struct ib_uverbs_create_qp_resp has a '__u32 driver_data[0]' aligned
> to 8 bytes thus the alignment of this struct can be 4 or 8.
>
> I generally don't recommend relying on this weird side effect, and encourage
> explicit padding when possible, but since the intent of this new driver is to be ABI
> compatible with the old driver, it should be kept the same.
>
> The userspace has a number of static_asserts which are designed to automatically
> check these various cases. I assume Intel has revised the userspace to use the
> new struct names and tested it..
>
Thanks Jason for the explanation! Yes these abi structs are kept the same for old user-space compatibility. And yes its been tested with old user-space.
next prev parent reply other threads:[~2020-05-27 1:58 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-20 7:03 [RDMA RFC v6 00/16] Intel RDMA Driver Updates 2020-05-19 Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 01/16] RDMA/irdma: Add driver framework definitions Jeff Kirsher
2020-05-20 7:26 ` Greg KH
2020-05-27 1:57 ` Saleem, Shiraz
2020-05-20 7:04 ` [RDMA RFC v6 02/16] RDMA/irdma: Implement device initialization definitions Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 03/16] RDMA/irdma: Implement HW Admin Queue OPs Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 04/16] RDMA/irdma: Add HMC backing store setup functions Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 05/16] RDMA/irdma: Add privileged UDA queue implementation Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 06/16] RDMA/irdma: Add QoS definitions Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 07/16] RDMA/irdma: Add connection manager Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 08/16] RDMA/irdma: Add PBLE resource manager Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 09/16] RDMA/irdma: Implement device supported verb APIs Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 10/16] RDMA/irdma: Add RoCEv2 UD OP support Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 11/16] RDMA/irdma: Add user/kernel shared libraries Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 12/16] RDMA/irdma: Add miscellaneous utility definitions Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 13/16] RDMA/irdma: Add dynamic tracing for CM Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 14/16] RDMA/irdma: Add ABI definitions Jeff Kirsher
2020-05-20 7:54 ` Gal Pressman
2020-05-20 8:52 ` Greg KH
2020-05-20 9:02 ` Gal Pressman
2020-05-20 12:37 ` Jason Gunthorpe
2020-05-27 1:58 ` Saleem, Shiraz [this message]
2020-05-20 7:04 ` [RDMA RFC v6 15/16] RDMA/irdma: Add irdma Kconfig/Makefile and remove i40iw Jeff Kirsher
2020-05-20 7:04 ` [RDMA RFC v6 16/16] RDMA/irdma: Update MAINTAINERS file Jeff Kirsher
2020-05-20 7:49 ` Gal Pressman
2020-05-27 1:58 ` Saleem, Shiraz
2020-05-21 14:12 ` [RDMA RFC v6 00/16] Intel RDMA Driver Updates 2020-05-19 Jason Gunthorpe
2020-05-27 1:58 ` Saleem, Shiraz
2020-05-27 5:08 ` Leon Romanovsky
2020-05-29 15:21 ` Saleem, Shiraz
2020-06-01 14:28 ` Jason Gunthorpe
2020-06-02 22:59 ` Saleem, Shiraz
2020-06-02 23:29 ` 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=9DD61F30A802C4429A01CA4200E302A7EE04048D@fmsmsx124.amr.corp.intel.com \
--to=shiraz.saleem@intel.com \
--cc=davem@davemloft.net \
--cc=dledford@redhat.com \
--cc=galpress@amazon.com \
--cc=gregkh@linuxfoundation.org \
--cc=jeffrey.t.kirsher@intel.com \
--cc=jgg@mellanox.com \
--cc=linux-rdma@vger.kernel.org \
--cc=mustafa.ismail@intel.com \
--cc=netdev@vger.kernel.org \
--cc=nhorman@redhat.com \
--cc=poswald@suse.com \
--cc=sassmann@redhat.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).