> -----Original Message----- > From: Joyce Kong [mailto:joyce.k...@arm.com] > Sent: Wednesday, October 23, 2019 4:55 AM > > There are a lot functions of bit operations scattered and > duplicated in PMDs, consolidating them into a common API > family is necessary. Furthermore, the bit operation is > mostly applied to the IO devices, so use __ATOMIC_ACQ_REL > to ensure the ordering. > > Signed-off-by: Joyce Kong <joyce.k...@arm.com> > Reviewed-by: Gavin Hu <gavin...@arm.com> > --- > lib/librte_eal/common/Makefile | 1 + > lib/librte_eal/common/include/rte_io_bitops.h | 112 > ++++++++++++++++++++++++++ > lib/librte_eal/common/meson.build | 1 + > 3 files changed, 114 insertions(+) > create mode 100644 lib/librte_eal/common/include/rte_io_bitops.h > > diff --git a/lib/librte_eal/common/Makefile > b/lib/librte_eal/common/Makefile > index a00d4fc..3831313 100644 > --- a/lib/librte_eal/common/Makefile > +++ b/lib/librte_eal/common/Makefile > @@ -18,6 +18,7 @@ INC += rte_malloc.h rte_keepalive.h rte_time.h > INC += rte_service.h rte_service_component.h > INC += rte_bitmap.h rte_vfio.h rte_hypervisor.h rte_test.h > INC += rte_reciprocal.h rte_fbarray.h rte_uuid.h > +INC += rte_io_bitops.h > > GENERIC_INC := rte_atomic.h rte_byteorder.h rte_cycles.h > rte_prefetch.h > GENERIC_INC += rte_memcpy.h rte_cpuflags.h > diff --git a/lib/librte_eal/common/include/rte_io_bitops.h > b/lib/librte_eal/common/include/rte_io_bitops.h > new file mode 100644 > index 0000000..5f778b8 > --- /dev/null > +++ b/lib/librte_eal/common/include/rte_io_bitops.h > @@ -0,0 +1,112 @@ > +/* SPDX-License-Identifier: BSD-3-Clause > + * Copyright(c) 2019 Arm Limited > + */ > + > +#ifndef _RTE_IO_BITOPS_H_ > +#define _RTE_IO_BITOPS_H_ > + > +/** > + * @file > + * Bit Operations > + * > + * This file defines a generic API for bit operations. > + */
-> This file defines a generic API for I/O device bit operations. > + > +#include <rte_lcore.h> This library doesn't do any lcore operations. Please use the appropriate headers. > + > +/** > + * Get a bit. > + * > + * @param nr > + * The bit to get. > + * @param addr > + * The address to count from. The address to count from. -> The address holding the bit. (Applies to all functions.) > + * @return > + * The value of the bit. The description of the return value can be misinterpreted. The return value is not the value of the bit, which is 0 or 1, but the value of the word holding the bit, masked with the bit position. (Applies to all functions returning a value.) > + */ > +static inline int32_t > +rte_io_get_bit(uint32_t nr, uint64_t *addr) The return type should be an unsigned type. (Applies to all functions returning a value.) The addr type for 32 bit operations should not be uint64_t *, but uint32_t *. The __atomic_xxx functions actually use this type for something, and I think it would access the wrong 32 bits on a big endian CPU. In some of the functions below, nr is signed (int32_t); it should be unsigned (uint32_t). And a suggestion: Consider changing the type of nr from uint32_t to unsigned int. > +{ > + return __atomic_load_n(addr, __ATOMIC_ACQUIRE) & (1UL << nr); > +} > + > +/** > + * Set a bit to 1. > + * > + * @param nr > + * The bit to set. > + * @param addr > + * The address to count from. > + */ > +static inline void > +rte_io_set_bit(uint32_t nr, uint64_t *addr) > +{ > + __atomic_fetch_or(addr, (1UL << nr), __ATOMIC_ACQ_REL); > +} > + > +/** > + * Set a bit to 0. > + * > + * @param nr > + * The bit to set. > + * @param addr > + * The address to count from. > + */ > +static inline void > +rte_io_clear_bit(int32_t nr, uint64_t *addr) > +{ > + __atomic_fetch_and(addr, ~(1UL << nr), __ATOMIC_ACQ_REL); > +} > + > +/** > + * Test if a bit is 1. > + * > + * @param nr > + * The bit to test. > + * @param addr > + * The address to count from. > + * @return > + * 1 if the bit is 1; else 0. > + */ > +static inline int32_t > +rte_io_test_bit(int32_t nr, uint64_t *addr) > +{ > + return (__atomic_load_n(addr, __ATOMIC_ACQUIRE) & (1UL << nr)) != > 0; > +} All the other functions in this library return a word with the bit masked. This function returns 0 or 1. I think it should return a word value, similar to the other functions. > + > +/** > + * Set a bit to 1 and return its old value. > + * > + * @param nr > + * The bit to set. > + * @param addr > + * The address to count from. > + * @return > + * The old value of the bit. > + */ > +static inline int32_t > +rte_io_test_and_set_bit(int32_t nr, uint64_t *addr) > +{ > + unsigned long mask = (1UL << nr); unsigned long mask -> uint32_t mask. (Also applies to other functions.) > + > + return __atomic_fetch_or(addr, mask, __ATOMIC_ACQ_REL) & mask; > +} > + > +/** > + * Set a bit to 0 and return its old value. > + * > + * @param nr > + * The bit to set. > + * @param addr > + * The address to count from. > + * @return > + * The old value of the bit. > + */ > +static inline int32_t > +rte_io_test_and_clear_bit(int32_t nr, uint64_t *addr) > +{ > + unsigned long mask = (1UL << nr); > + > + return __atomic_fetch_and(addr, ~mask, __ATOMIC_ACQ_REL) & mask; > +} > +#endif /* _RTE_IO_BITOPS_H_ */ > diff --git a/lib/librte_eal/common/meson.build > b/lib/librte_eal/common/meson.build > index 386577c..0a65d04 100644 > --- a/lib/librte_eal/common/meson.build > +++ b/lib/librte_eal/common/meson.build > @@ -52,6 +52,7 @@ common_headers = files( > 'include/rte_alarm.h', > 'include/rte_branch_prediction.h', > 'include/rte_bus.h', > + 'include/rte_io_bitops.h', > 'include/rte_bitmap.h', > 'include/rte_class.h', > 'include/rte_common.h', > -- > 2.7.4 >