public inbox for linux-rdma@vger.kernel.org
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Bart Van Assche <bvanassche@acm.org>
Cc: Leon Romanovsky <leonro@mellanox.com>,
	Doug Ledford <dledford@redhat.com>,
	linux-rdma@vger.kernel.org,
	Nicolas Morey-Chaisemartin <nmoreychaisemartin@suse.com>
Subject: Re: [PATCH 3/5] RDMA/srp: Apply the __packed attribute to members instead of structures
Date: Mon, 24 May 2021 20:00:38 -0300	[thread overview]
Message-ID: <20210524230038.GP1002214@nvidia.com> (raw)
In-Reply-To: <7a77c2ed-5336-11ea-162e-539b593a1ce7@acm.org>

On Sun, May 23, 2021 at 08:41:25PM -0700, Bart Van Assche wrote:
> On 5/20/21 7:48 AM, Jason Gunthorpe wrote:
> > On Tue, May 11, 2021 at 08:27:50PM -0700, Bart Van Assche wrote:
> >> @@ -107,10 +107,10 @@ struct srp_direct_buf {
> >>   * having the 20-byte structure padded to 24 bytes on 64-bit architectures.
> >>   */
> >>  struct srp_indirect_buf {
> >> -	struct srp_direct_buf	table_desc;
> >> +	struct srp_direct_buf	table_desc __packed;
> >>  	__be32			len;
> >> -	struct srp_direct_buf	desc_list[];
> >> -} __attribute__((packed));
> >> +	struct srp_direct_buf	desc_list[] __packed;
> >> +};
> >>  
> >>  /* Immediate data buffer descriptor as defined in SRP2. */
> >>  struct srp_imm_buf {
> >> @@ -175,13 +175,13 @@ struct srp_login_rsp {
> >>  	u8	opcode;
> >>  	u8	reserved1[3];
> >>  	__be32	req_lim_delta;
> >> -	u64	tag;
> >> +	u64	tag __packed;
> > 
> > What you really want is just something like this:
> > 
> > typedef u64 __attribute__((aligned(4))) compat_u64;
> > 
> > And then use that every place you have the u64 and forget about packed
> > entirely.
> 
> Really? My understanding is that the aligned attribute can be used to
> increase alignment of a structure member but not to decrease it. 

I didn't test it, but that is pre-existing code in Linux.. Maybe it
doesn't work and/or needs packed too

> Additionally, there is a recommendation in
> Documentation/process/coding-style.rst not to introduce new typedefs.

That we tend to selective ignore <shrug>

> > Except for a couple exceptions IBA mads are always aligned to 4 bytes,
> > only the 64 bit quantities are unaligned.
> > 
> > But really this whole thing should be replaced with the IBA_FIELD
> > macros like include/rdma/ibta_vol1_c12.h demos.
> > 
> > Then it would be sparse safe and obviously endian correct as well.
> 
> I prefer a structure over the IBA_FIELD macros so I will change __packed
> into __packed __aligned(4).

IMHO the struct is alot worse due to lack of endian safety and
complexity in establishing the layout.

Jason

  reply	other threads:[~2021-05-24 23:00 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-12  3:27 [PATCH 0/5] SRP kernel patches for kernel v5.14 Bart Van Assche
2021-05-12  3:27 ` [PATCH 1/5] RDMA/ib_hdrs.h: Remove a superfluous cast Bart Van Assche
2021-05-19  9:00   ` Leon Romanovsky
2021-05-24  3:12     ` Bart Van Assche
2021-05-12  3:27 ` [PATCH 2/5] RDMA/srp: Add more structure size checks Bart Van Assche
2021-05-19  9:01   ` Leon Romanovsky
2021-05-12  3:27 ` [PATCH 3/5] RDMA/srp: Apply the __packed attribute to members instead of structures Bart Van Assche
2021-05-20 14:48   ` Jason Gunthorpe
2021-05-24  3:41     ` Bart Van Assche
2021-05-24 23:00       ` Jason Gunthorpe [this message]
2021-05-12  3:27 ` [PATCH 4/5] RDMA/srp: Fix a recently introduced memory leak Bart Van Assche
2021-05-12  8:15   ` Max Gurtovoy
2021-05-12 16:14     ` Bart Van Assche
2021-05-12  3:27 ` [PATCH 5/5] RDMA/srp: Make struct scsi_cmnd and struct srp_request adjacent Bart Van Assche
2021-05-19  9:09   ` Leon Romanovsky
2021-05-19 15:13     ` Bart Van Assche
2021-05-19 15:32       ` Leon Romanovsky

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=20210524230038.GP1002214@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=bvanassche@acm.org \
    --cc=dledford@redhat.com \
    --cc=leonro@mellanox.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=nmoreychaisemartin@suse.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