DPDK patches and discussions
 help / color / mirror / Atom feed
From: Ferruh Yigit <ferruh.yigit@amd.com>
To: "Morten Brørup" <mb@smartsharesystems.com>,
	dev@dpdk.org, roretzla@linux.microsoft.com, rmody@marvell.com,
	timothy.mcdaniel@intel.com, matan@nvidia.com,
	viacheslavo@nvidia.com
Cc: ruifeng.wang@arm.com, zhoumin@loongson.cn,
	drc@linux.vnet.ibm.com, kda@semihalf.com,
	bruce.richardson@intel.com, konstantin.v.ananyev@yandex.ru
Subject: Re: [PATCH v7 4/4] eal: add nonnull and access function attributes
Date: Mon, 16 Jan 2023 17:02:06 +0000	[thread overview]
Message-ID: <b1930d00-4e9c-00d9-5f49-d8fe2c559e8f@amd.com> (raw)
In-Reply-To: <20230116130724.50277-4-mb@smartsharesystems.com>

On 1/16/2023 1:07 PM, Morten Brørup wrote:
> Add nonnull function attribute to help the compiler detect a NULL
> pointer being passed to a function not accepting NULL pointers as an
> argument at build time.
> 
> Add access function attributes to tell the compiler how a function
> accesses memory pointed to by its pointer arguments.
> 
> Add these attributes to the rte_memcpy() function, as the first in
> hopefully many to come.
> 
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> Acked-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
> Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
> ---
> v7:
> * Fix indentation. (checkpatch)
> v6:
> * Make this the last patch in the series, putting the fixes first.
>   (David)
> * Split the generic access macro into dedicated macros per access mode.
>   (David)
> v5:
> * No changes.
> v4:
> * No changes.
> v3:
> * No changes.
> v2:
> * Only define "nonnull" for GCC and CLANG.
> * Append _param/_params to prepare for possible future attributes
>   attached directly to the individual parameters, like __rte_unused.
> * Use RTE_TOOLCHAIN_GCC instead of RTE_CC_GCC, to fix complaints about
>   GCC_VERSION being undefined.
> * Try to fix Doxygen compliants.
> ---
>  lib/eal/arm/include/rte_memcpy_32.h |  8 +++
>  lib/eal/arm/include/rte_memcpy_64.h |  6 +++
>  lib/eal/include/rte_common.h        | 76 +++++++++++++++++++++++++++++
>  lib/eal/ppc/include/rte_memcpy.h    |  3 ++
>  lib/eal/x86/include/rte_memcpy.h    |  6 +++
>  5 files changed, 99 insertions(+)
> 
> diff --git a/lib/eal/arm/include/rte_memcpy_32.h b/lib/eal/arm/include/rte_memcpy_32.h
> index fb3245b59c..a625a91951 100644
> --- a/lib/eal/arm/include/rte_memcpy_32.h
> +++ b/lib/eal/arm/include/rte_memcpy_32.h
> @@ -14,6 +14,8 @@ extern "C" {
>  
>  #include "generic/rte_memcpy.h"
>  
> +#include <rte_common.h>
> +
>  #ifdef RTE_ARCH_ARM_NEON_MEMCPY
>  
>  #ifndef __ARM_NEON
> @@ -125,6 +127,9 @@ rte_mov256(uint8_t *dst, const uint8_t *src)
>  	memcpy((dst), (src), (n)) :          \
>  	rte_memcpy_func((dst), (src), (n)); })
>  
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static inline void *
>  rte_memcpy_func(void *dst, const void *src, size_t n)
>  {
> @@ -290,6 +295,9 @@ rte_mov256(uint8_t *dst, const uint8_t *src)
>  	memcpy(dst, src, 256);
>  }
>  
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static inline void *
>  rte_memcpy(void *dst, const void *src, size_t n)
>  {
> diff --git a/lib/eal/arm/include/rte_memcpy_64.h b/lib/eal/arm/include/rte_memcpy_64.h
> index 85ad587bd3..0c86237cc9 100644
> --- a/lib/eal/arm/include/rte_memcpy_64.h
> +++ b/lib/eal/arm/include/rte_memcpy_64.h
> @@ -282,6 +282,9 @@ void rte_memcpy_ge64(uint8_t *dst, const uint8_t *src, size_t n)
>  }
>  
>  #if RTE_CACHE_LINE_SIZE >= 128
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static __rte_always_inline
>  void *rte_memcpy(void *dst, const void *src, size_t n)
>  {
> @@ -303,6 +306,9 @@ void *rte_memcpy(void *dst, const void *src, size_t n)
>  }
>  
>  #else
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static __rte_always_inline
>  void *rte_memcpy(void *dst, const void *src, size_t n)
>  {
> diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
> index 15765b408d..c9cd2c7496 100644
> --- a/lib/eal/include/rte_common.h
> +++ b/lib/eal/include/rte_common.h
> @@ -149,6 +149,82 @@ typedef uint16_t unaligned_uint16_t;
>  	__attribute__((format(printf, format_index, first_arg)))
>  #endif
>  
> +/**
> + * Tells compiler that the pointer arguments must be non-null.
> + *
> + * @param ...
> + *    Comma separated list of parameter indexes of pointer arguments.
> + */
> +#if defined(RTE_CC_GCC) || defined(RTE_CC_CLANG)
> +#define __rte_nonnull_params(...) \
> +	__attribute__((nonnull(__VA_ARGS__)))
> +#else
> +#define __rte_nonnull_params(...)
> +#endif
> +

What do you think to have a namespace for macros like '__rte_param_xxx',
so name macros as:
__rte_param_nonull
__rte_param_read_only
__rte_param_write_only

No strong opinion, it just feels tidier this way

> +/**
> + * Tells compiler that the memory pointed to by a pointer argument
> + * is only read, not written to, by the function.
> + * It might not be accessed at all.
> + *
> + * @param ...
> + *    Parameter index of pointer argument.
> + *    Optionally followeded by comma and parameter index of size argument.

s/followeded/followed/

multiple occurrences below

> + */
> +#if defined(RTE_TOOLCHAIN_GCC) && (GCC_VERSION >= 100400)
> +#define __rte_read_only_param(...) \
> +	__attribute__((access(read_only, __VA_ARGS__)))
> +#else
> +#define __rte_read_only_param(...)
> +#endif
> +
> +/**
> + * Tells compiler that the memory pointed to by a pointer argument
> + * is only written to, not read, by the function.
> + * It might not be accessed at all.
> + *
> + * @param ...
> + *    Parameter index of pointer argument.
> + *    Optionally followeded by comma and parameter index of size argument.
> + */
> +#if defined(RTE_TOOLCHAIN_GCC) && (GCC_VERSION >= 100400)
> +#define __rte_write_only_param(...) \
> +	__attribute__((access(write_only, __VA_ARGS__)))
> +#else
> +#define __rte_write_only_param(...)
> +#endif
> +
> +/**
> + * Tells compiler that the memory pointed to by a pointer argument
> + * is both read and written to by the function.
> + * It might not be read, written to, or accessed at all.
> + *

What I understand is difference between 'read_write' access mode and
'nonnull' attribute is,
'read_write' access mode additionally checks that variable is
initialized before used in this function.
Should we document this detail, although it is gcc manual I think it
helps to clarify here.

> + * @param ...
> + *    Parameter index of pointer argument.
> + *    Optionally followeded by comma and parameter index of size argument.
> + */
> +#if defined(RTE_TOOLCHAIN_GCC) && (GCC_VERSION >= 100400)
> +#define __rte_read_write_param(...) \
> +	__attribute__((access(read_write, __VA_ARGS__)))
> +#else
> +#define __rte_read_write_param(...)
> +#endif
> +
> +/**
> + * Tells compiler that the memory pointed to by a pointer argument
> + * is not accessed by the function.
> + *
> + * @param ...
> + *    Parameter index of pointer argument.
> + *    Optionally followeded by comma and parameter index of size argument.
> + */
> +#if defined(RTE_TOOLCHAIN_GCC) && (GCC_VERSION >= 100400)
> +#define __rte_no_access_param(...) \
> +	__attribute__((access(none, __VA_ARGS__)))
> +#else
> +#define __rte_no_access_param(...)
> +#endif
> +
>  /**
>   * Tells compiler that the function returns a value that points to
>   * memory, where the size is given by the one or two arguments.
> diff --git a/lib/eal/ppc/include/rte_memcpy.h b/lib/eal/ppc/include/rte_memcpy.h
> index 6f388c0234..c36869a414 100644
> --- a/lib/eal/ppc/include/rte_memcpy.h
> +++ b/lib/eal/ppc/include/rte_memcpy.h
> @@ -84,6 +84,9 @@ rte_mov256(uint8_t *dst, const uint8_t *src)
>  	memcpy((dst), (src), (n)) :          \
>  	rte_memcpy_func((dst), (src), (n)); })
>  
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static inline void *
>  rte_memcpy_func(void *dst, const void *src, size_t n)
>  {
> diff --git a/lib/eal/x86/include/rte_memcpy.h b/lib/eal/x86/include/rte_memcpy.h
> index d4d7a5cfc8..ee543aa37d 100644
> --- a/lib/eal/x86/include/rte_memcpy.h
> +++ b/lib/eal/x86/include/rte_memcpy.h
> @@ -42,6 +42,9 @@ extern "C" {
>   * @return
>   *   Pointer to the destination data.
>   */
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static __rte_always_inline void *
>  rte_memcpy(void *dst, const void *src, size_t n);
>  
> @@ -859,6 +862,9 @@ rte_memcpy_aligned(void *dst, const void *src, size_t n)
>  	return ret;
>  }
>  
> +__rte_nonnull_params(1, 2)
> +__rte_write_only_param(1, 3)
> +__rte_read_only_param(2, 3)
>  static __rte_always_inline void *
>  rte_memcpy(void *dst, const void *src, size_t n)
>  {


  reply	other threads:[~2023-01-16 17:02 UTC|newest]

Thread overview: 77+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-12-02 15:34 [PATCH] " Morten Brørup
2022-12-02 20:02 ` Tyler Retzlaff
2022-12-03 14:22 ` [PATCH v2] " Morten Brørup
2022-12-05 10:17   ` Ruifeng Wang
2022-12-12  7:40   ` Morten Brørup
2022-12-28 10:27 ` [PATCH v3 1/2] " Morten Brørup
2022-12-28 10:27   ` [PATCH v3 2/2] net/bnx2x: fix warnings about rte_memcopy lengths Morten Brørup
2022-12-28 11:40 ` [PATCH v4 1/2] eal: add nonnull and access function attributes Morten Brørup
2022-12-28 11:40   ` [PATCH v4 2/2] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2022-12-28 15:10 ` [PATCH v5 1/4] eal: add nonnull and access function attributes Morten Brørup
2022-12-28 15:10   ` [PATCH v5 2/4] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2022-12-28 16:13     ` Stanisław Kardach
2022-12-28 16:38       ` Morten Brørup
2022-12-28 17:03         ` Stephen Hemminger
2022-12-28 17:37           ` Morten Brørup
2023-01-09 10:36             ` David Marchand
2022-12-28 15:10   ` [PATCH v5 3/4] event/dlb2: remove superfluous rte_memcpy Morten Brørup
2022-12-28 17:01     ` Stephen Hemminger
2022-12-28 15:10   ` [PATCH v5 4/4] net/mlx5: fix warning about rte_memcpy length Morten Brørup
2023-01-09 11:08   ` [PATCH v5 1/4] eal: add nonnull and access function attributes Thomas Monjalon
2023-01-09 12:16     ` Morten Brørup
2023-01-09 11:22   ` David Marchand
2023-01-09 12:28     ` Morten Brørup
2023-01-16 12:49     ` Morten Brørup
2023-01-16 12:44 ` [PATCH v6 1/4] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2023-01-16 12:44   ` [PATCH v6 2/4] event/dlb2: remove superfluous rte_memcpy Morten Brørup
2023-01-16 12:44   ` [PATCH v6 3/4] net/mlx5: fix warning about rte_memcpy length Morten Brørup
2023-01-16 12:44   ` [PATCH v6 4/4] eal: add nonnull and access function attributes Morten Brørup
2023-01-16 13:07 ` [PATCH v7 1/4] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2023-01-16 13:07   ` [PATCH v7 2/4] event/dlb2: remove superfluous rte_memcpy Morten Brørup
2023-02-09 16:51     ` Morten Brørup
2023-02-09 18:50       ` Sevincer, Abdullah
2023-02-10  7:43         ` Morten Brørup
2024-02-23 13:19           ` Jerin Jacob
2024-02-23 13:49             ` [PATCH v8] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2024-02-23 14:00             ` [PATCH v9] " Morten Brørup
2024-02-26  8:34               ` Jerin Jacob
2024-02-26 14:47                 ` Morten Brørup
2024-02-27 11:00                   ` Jerin Jacob
2024-02-27 11:27                     ` Morten Brørup
2024-02-27 19:06                       ` Stephen Hemminger
2024-02-28  9:02                         ` Bruce Richardson
2024-02-28  8:27                       ` Raslan Darawsheh
2024-02-26 21:45               ` Stephen Hemminger
2023-01-16 13:07   ` [PATCH v7 3/4] net/mlx5: fix warning about rte_memcpy length Morten Brørup
2023-02-09 16:54     ` Morten Brørup
2023-02-10  9:13       ` Slava Ovsiienko
2024-03-13 10:00     ` Raslan Darawsheh
2023-01-16 13:07   ` [PATCH v7 4/4] eal: add nonnull and access function attributes Morten Brørup
2023-01-16 17:02     ` Ferruh Yigit [this message]
2023-01-17  8:19       ` Morten Brørup
2023-01-17 21:16         ` Tyler Retzlaff
2023-01-18  8:31           ` Morten Brørup
2023-01-18 17:23             ` Stephen Hemminger
2023-01-31 11:14             ` David Marchand
2023-01-31 12:23               ` Morten Brørup
2023-01-31 18:26                 ` Tyler Retzlaff
2023-01-31 22:52                   ` Thomas Monjalon
2023-02-01 12:50                     ` Morten Brørup
2023-02-01 13:15                       ` Thomas Monjalon
2023-02-06 16:11                         ` David Marchand
2023-02-06 16:49                           ` Morten Brørup
2023-02-06 17:28                             ` Tyler Retzlaff
2023-05-08 12:32                               ` Morten Brørup
2023-10-13 23:31                                 ` Morten Brørup
2023-04-04 13:41     ` Morten Brørup
2023-04-04 13:51       ` David Marchand
2023-04-04 14:01         ` Morten Brørup
2023-04-04 14:19           ` David Marchand
2023-02-09 16:49   ` [PATCH v7 1/4] net/bnx2x: fix warnings about rte_memcpy lengths Morten Brørup
2023-02-26  9:21     ` Marvell QLogic bnx2x PMD support status Morten Brørup
2023-03-09 10:25     ` [PATCH v7 1/4] net/bnx2x: fix warnings about rte_memcpy lengths David Marchand
2023-03-09 10:33       ` Thomas Monjalon
2023-03-09 16:11         ` [EXT] " Akhil Goyal
     [not found]           ` <SJ0PR18MB390039A8C34E485F8D2D518EA1B59@SJ0PR18MB3900.namprd18.prod.outlook.com>
2023-03-09 16:19             ` Devendra Singh Rawat
2023-03-09 16:23     ` Stephen Hemminger
2024-02-23 12:39       ` Jerin Jacob

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=b1930d00-4e9c-00d9-5f49-d8fe2c559e8f@amd.com \
    --to=ferruh.yigit@amd.com \
    --cc=bruce.richardson@intel.com \
    --cc=dev@dpdk.org \
    --cc=drc@linux.vnet.ibm.com \
    --cc=kda@semihalf.com \
    --cc=konstantin.v.ananyev@yandex.ru \
    --cc=matan@nvidia.com \
    --cc=mb@smartsharesystems.com \
    --cc=rmody@marvell.com \
    --cc=roretzla@linux.microsoft.com \
    --cc=ruifeng.wang@arm.com \
    --cc=timothy.mcdaniel@intel.com \
    --cc=viacheslavo@nvidia.com \
    --cc=zhoumin@loongson.cn \
    /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).