| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20080620081109.958a5d6f@mailhost.serverengines.com>
Date: Fri, 20 Jun 2008 01:11:09 -0700
From: "Subbu Seetharaman" <subbus@...verengines.com>
To: "Christoph Hellwig" <hch@...radead.org>
Cc: jeff@...zik.org, netdev@...r.kernel.org
Subject: Re: [PATCH 4/12] benet: beclib (h/w access lib) header files
Christoph,
All the stubs and wrappers will be removed.
We will also use the native Linux mechanism
where ever you have pointed out and
make beclib it's own module.
Can you point to the bitmask infrastructure
in Linux that you are referring to below ?
If they can eliminate or simplify the
amap macros we will do that.
Thanks.
Subbu
> +static inline u32
> +amap_get(void *ptr, u32 dw_offset, u32 mask, u32 offset)
> +{
> + u32 *dw = (u32 *)ptr;
> + return mask & (*(dw + dw_offset) >> offset);
> +}
> +#define AMAP_GET_BITS_PTR(_struct_, _register_, _structPtr_) \
> + amap_get(_structPtr_, AMAP_WORD_OFFSET(_struct_, _register_), \
> + AMAP_BIT_MASK(_struct_, _register_), \
> + AMAP_BIT_OFFSET(_struct_, _register_)) \
Why do you need all this infrastructure? Any reason the regular
bitmasks in Linux are not enough?
----- Original Message -----
From: Christoph Hellwig [mailto:hch@...radead.org]
To: Subbu Seetharaman [mailto:subbus@...verengines.com]
Cc: jeff@...zik.org, netdev@...r.kernel.org
Sent: Thu, 19 Jun 2008 03:24:32 -0700
Subject: Re: [PATCH 4/12] benet: beclib (h/w access lib) header files
> +#define TRUE (1)
> +#define FALSE (0)
Please use the standard true/false as provided by the kernel headers.
> +#if !defined(SG_PACK)
> +#define SG_PACK __attribute__ ((packed))
> +#endif
Please use __packed
> +/*
> + *--------- Assert related --------
> + * a way to force compilation error
> + */
> +
> +#define SA_C_ASSERT(_expression_) { \
> + struct __COMPILE_TIME_ASSERT__ { \
> + u8 __COMPILE_TIME_ASSERT__[(_expression_)?1: -1]; \
> + }; \
> +}
Please use BUG_ON.
> +#define SA_COMPILETIME_NAMED_TYPEDEF_ASSERT(_name_, _expression_) \
> + struct _name_##_COMPILE_TIME_ASSERT_ { \
> + u8 _name##_COMPILE_TIME_ASSERT_[(_expression_)?1: -1]; \
> + };
Please use BUILD_BUG_ON.
> +static inline u32 sa_required_bits(u32 x)
> + * Returns the log base 2 for a number, always rounding down.
> + * sa_log2 example input->output
> + * 0->4294967295, 1->0, 2->1, 3->1, 4->2, 5->2, 6->2, 7->2, 8->3, 9->3
> + */
> +static inline u32 sa_log2(u32 x)
> +{
> + return sa_required_bits(x) - 1;
Please take a look at include/linux/log2.h
> + * Status code datatype
> + */
> +typedef int SA_STATUS, *PSA_STATUS;
Please kill this and just use normal errno values.
> +#define SA_PTR_TO_INT(_p_) ((size_t)(_p_))
> +#define SA_PTR_TO_U32(_p_) ((u32)SA_PTR_TO_INT(_p_))
> +#define SA_PTR_TO_U64(_p_) ((u64)SA_PTR_TO_INT(_p_))
> +
> +#define SA_PTR_TO_INT(_p_) ((size_t)(_p_))
> +#define SA_U64_TO_PTR(_u_) ((void *)(size_t)(_u_))
> +#define SA_U32_TO_PTR(_u_) ((void *)(size_t)(_u_))
These warnings are there for a reason, and having the casts spelled
out where they are required is a good warnings sign.
> +#define SA_PAGE_OFFSET(_addr_) (SA_PTR_TO_INT(_addr_) & (PAGE_SIZE-1))
Should be lower case and without that sa prefix.
> +/* Returns byte offset of field within given struct type. */
> +#define SA_FIELD_OFFSET(_type_, _field_) \
> + (SA_PTR_TO_U32((u8 *)&(((struct _type_ *)0)->_field_)))
> +
> +/* Returns byte size of given field with a structure. */
> +#define SA_SIZEOF_FIELD(_type_, _field_) \
> + sizeof(((struct _type_ *)0)->_field_)
> +
> +/* Returns the number of items in the field array. */
> +#define SA_NUMBER_OF_FIELD(_type_, _field_) \
> + (SA_SIZEOF_FIELD(_type_, _field_) / \
> + sizeof((((struct _type_ *)0)->_field_[0])))
> +
Please use the kernel.h helpers.
> +/* descriptor for a physically contiguous memory region */
> +struct sa_sgl {
> + u32 length;
> + void *va;
> + u64 pa;
> +} ;
What do you actually needs this an the helpers for? It seems a rather
odd data structure because linux scatter/gather lists are page-based.
> +/*----------- amap bit filed get / set macros and functions -----*/
> +/*
> + * Structures defined in the map header files (under fw/amap/) with names
> + * in the format BE_<name>_AMAP are pseudo structures with members
> + * of type BE_BIT. These structures are templates that are used in
> + * conjuntion with the structures with names in the format
> + * <name>_AMAP to calculate the bit masks and bit offsets to get or set
> + * bit fields in structures. The structures <name>_AMAP are arrays
> + * of 32 bits words and have the correct size. The following macros
> + * provide convenient ways to get and set the various members
> + * in the structures without using strucctures with bit fields.
> + * Always use the macros AMAP_GET_BITS_PTR and AMAP_SET_BITS_PTR
> + * macros to extract and set various members.
> + */
> +/* Each bit is 1 byte in array maps. */
> +typedef u8 AMAP_BIT, BE_BIT;
Please just u8 instead of these ugly typedefs.
> +static inline u32
> +amap_get(void *ptr, u32 dw_offset, u32 mask, u32 offset)
> +{
> + u32 *dw = (u32 *)ptr;
> + return mask & (*(dw + dw_offset) >> offset);
> +}
> +#define AMAP_GET_BITS_PTR(_struct_, _register_, _structPtr_) \
> + amap_get(_structPtr_, AMAP_WORD_OFFSET(_struct_, _register_), \
> + AMAP_BIT_MASK(_struct_, _register_), \
> + AMAP_BIT_OFFSET(_struct_, _register_)) \
Why do you need all this infrastructure? Any reason the regular
bitmasks in Linux are not enough?
> +#if !defined(BE_CONFIG)
> +#define BE_CONFIG 0
> +#define CONFIG0
> +#endif
What's the reasons for this?
> +/* Callback types. Deprecate most of these. */
> +typedef void (*MCC_WRB_CQE_CALLBACK) (void *context,
> + BESTATUS status,
> + struct MCC_WRB_AMAP *optional_wrb);
> +typedef void (*MCC_ASYNC_EVENT_CALLBACK) (void *context,
> + u32 event_code,
> + void *event);
> +
> +typedef BESTATUS(*EQ_CALLBACK) (struct be_function_object *,
> + struct be_eq_object *, void *context);
> +typedef BESTATUS(*CQ_CALLBACK) (struct be_function_object *,
> + struct be_cq_object *, void *context);
Lowercase, please.
> +/* --- BE_FUNCTION_ENUM --- */
> +#define BE_FUNCTION_TYPE_ISCSI (0)
> +#define BE_FUNCTION_TYPE_NETWORK (1)
> +#define BE_FUNCTION_TYPE_ARM (2)
> +
> +enum BE_FUNCTION_ENUM {
> + ENUM_BE_FUNCTION_TYPE_ISCSI = 0x0UL,
> + ENUM_BE_FUNCTION_TYPE_NETWORK = 0x1UL,
> + ENUM_BE_FUNCTION_TYPE_ARM = 0x2UL
> +} ;
Why is this done twice? Also the enum as every type name should be
lower cased.
___________________________________________________________________________________
This message, together with any attachment(s), contains confidential and proprietary information of
ServerEngines Corporation and is intended only for the designated recipient(s) named above. Any unauthorized
review, printing, retention, copying, disclosure or distribution is strictly prohibited. If you are not the
intended recipient of this message, please immediately advise the sender by reply email message and
delete all copies of this message and any attachment(s). Thank you.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists