From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id E614968CE for ; Thu, 9 Oct 2014 07:54:37 +0200 (CEST) Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by fmsmga101.fm.intel.com with ESMTP; 08 Oct 2014 23:01:56 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.04,683,1406617200"; d="scan'208";a="611733004" Received: from fmsmsx104.amr.corp.intel.com ([10.18.124.202]) by fmsmga002.fm.intel.com with ESMTP; 08 Oct 2014 23:01:55 -0700 Received: from fmsmsx152.amr.corp.intel.com (10.18.125.5) by fmsmsx104.amr.corp.intel.com (10.18.124.202) with Microsoft SMTP Server (TLS) id 14.3.195.1; Wed, 8 Oct 2014 23:01:55 -0700 Received: from shsmsx151.ccr.corp.intel.com (10.239.6.50) by FMSMSX152.amr.corp.intel.com (10.18.125.5) with Microsoft SMTP Server (TLS) id 14.3.195.1; Wed, 8 Oct 2014 23:01:54 -0700 Received: from shsmsx104.ccr.corp.intel.com ([169.254.5.230]) by SHSMSX151.ccr.corp.intel.com ([169.254.3.172]) with mapi id 14.03.0195.001; Thu, 9 Oct 2014 14:01:53 +0800 From: "Zhang, Helin" To: Marc Sune , "dev@dpdk.org" Thread-Topic: [dpdk-dev] [PATCH] KNI: use a memzone pool for KNI alloc/release Thread-Index: AQHP285+iUgjNWB580Cf22Ir3DYXhpwnUh3w Date: Thu, 9 Oct 2014 06:01:52 +0000 Message-ID: References: <1411985756-2744-1-git-send-email-marc.sune@bisdn.de> In-Reply-To: <1411985756-2744-1-git-send-email-marc.sune@bisdn.de> Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [10.239.127.40] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH] KNI: use a memzone pool for KNI alloc/release X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 09 Oct 2014 05:54:38 -0000 Hi Marc Thanks for the idea on KNI! I have comments/questions as follows. Please co= rrect me if I am wrong! > -----Original Message----- > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Marc Sune > Sent: Monday, September 29, 2014 6:16 PM > To: dev@dpdk.org > Subject: [dpdk-dev] [PATCH] KNI: use a memzone pool for KNI alloc/release >=20 > This patch implements the KNI memzone pool in order to: >=20 > * prevent memzone exhaustion when allocating/deallocating KNI > interfaces. What do you mean the "exhaustion"? Actually the memzones can be reused, tho= ugh they cannot be resized. > * be able to allocate KNI interfaces with the same name as > previously deallocated ones. I think the current implementation can already allocate the same name KNI i= nterface after being deallocated, as each time memzone reservation will try= to find the same name memzone first. If the name has already been used, th= at means it has even been allocated, we can find it back and reuse it, thou= gh the size cannot be changed. Have you encountered the problem you are trying to solve? Have you tried to= use current implementation of KNI for the case you are trying to fix for? If I am not wrong, we may not need the changes, as its current implementati= on already supports the scenarios you are trying to support. The only thing= is that the sizes of memzones cannot be changed, but it seems no idea on t= hat. >=20 > It adds a new API call, rte_kni_init(max_kni_ifaces) that shall be called= before > any call to rte_kni_alloc() if KNI is used. >=20 > Signed-off-by: Marc Sune > --- > lib/librte_kni/rte_kni.c | 302 > ++++++++++++++++++++++++++++++++++++++-------- > lib/librte_kni/rte_kni.h | 18 +++ > 2 files changed, 269 insertions(+), 51 deletions(-) >=20 > diff --git a/lib/librte_kni/rte_kni.c b/lib/librte_kni/rte_kni.c index > 76feef4..df55789 100644 > --- a/lib/librte_kni/rte_kni.c > +++ b/lib/librte_kni/rte_kni.c > @@ -40,6 +40,7 @@ > #include > #include >=20 > +#include > #include > #include > #include > @@ -58,7 +59,7 @@ >=20 > #define KNI_REQUEST_MBUF_NUM_MAX 32 >=20 > -#define KNI_MZ_CHECK(mz) do { if (mz) goto fail; } while (0) > +#define KNI_MEM_CHECK(cond) do { if (cond) goto kni_fail; } while (0) >=20 > /** > * KNI context > @@ -66,6 +67,7 @@ > struct rte_kni { > char name[RTE_KNI_NAMESIZE]; /**< KNI interface name */ > uint16_t group_id; /**< Group ID of KNI devices */ > + unsigned slot_id; /**< KNI pool slot ID */ > struct rte_mempool *pktmbuf_pool; /**< pkt mbuf mempool */ > unsigned mbuf_size; /**< mbuf size */ >=20 > @@ -88,10 +90,48 @@ enum kni_ops_status { > KNI_REQ_REGISTERED, > }; >=20 > +/** > +* KNI memzone pool slot > +*/ > +struct rte_kni_memzone_slot{ > + unsigned id; > + uint8_t in_use : 1; /**< slot in use */ > + > + //Memzones > + const struct rte_memzone *m_ctx; /**< KNI ctx */ > + const struct rte_memzone *m_tx_q; /**< TX queue */ > + const struct rte_memzone *m_rx_q; /**< RX queue */ > + const struct rte_memzone *m_alloc_q; /**< Allocated mbufs queue */ > + const struct rte_memzone *m_free_q; /**< To be freed mbufs queue > */ > + const struct rte_memzone *m_req_q; /**< Request queue */ > + const struct rte_memzone *m_resp_q; /**< Response queue */ > + const struct rte_memzone *m_sync_addr; > + > + /* Free linked list */ > + struct rte_kni_memzone_slot *next; /**< Next slot link.list */ > +}; > + > +/** > +* KNI memzone pool > +*/ > +struct rte_kni_memzone_pool{ > + uint8_t initialized : 1; /**< Global KNI pool init flag */ > + > + unsigned max_ifaces; /**< Max. num of KNI ifaces */ > + struct rte_kni_memzone_slot *slots; /**< Pool slots */ > + rte_spinlock_t mutex; /**< alloc/relase mutex */ > + > + //Free memzone slots linked-list > + struct rte_kni_memzone_slot *free; /**< First empty slot */ > + struct rte_kni_memzone_slot *free_tail; /**< Last empty slot */ > +}; > + > + > static void kni_free_mbufs(struct rte_kni *kni); static void > kni_allocate_mbufs(struct rte_kni *kni); >=20 > static volatile int kni_fd =3D -1; > +static struct rte_kni_memzone_pool kni_memzone_pool =3D {0}; >=20 > static const struct rte_memzone * > kni_memzone_reserve(const char *name, size_t len, int socket_id, @@ > -105,6 +145,154 @@ kni_memzone_reserve(const char *name, size_t len, int > socket_id, > return mz; > } >=20 > +/* Pool mgmt */ > +static struct rte_kni_memzone_slot* > +kni_memzone_pool_alloc(void) > +{ > + struct rte_kni_memzone_slot* slot; > + > + rte_spinlock_lock(&kni_memzone_pool.mutex); > + > + if(!kni_memzone_pool.free) { > + rte_spinlock_unlock(&kni_memzone_pool.mutex); > + return NULL; > + } > + > + slot =3D kni_memzone_pool.free; > + kni_memzone_pool.free =3D slot->next; > + > + if(!kni_memzone_pool.free) > + kni_memzone_pool.free_tail =3D NULL; > + > + rte_spinlock_unlock(&kni_memzone_pool.mutex); > + > + return slot; > +} > + > +static void > +kni_memzone_pool_dealloc(struct rte_kni_memzone_slot* slot) { > + rte_spinlock_lock(&kni_memzone_pool.mutex); > + > + if(kni_memzone_pool.free) > + kni_memzone_pool.free_tail->next =3D slot; > + else > + kni_memzone_pool.free =3D slot; > + > + kni_memzone_pool.free_tail =3D slot; > + slot->next =3D NULL; > + > + rte_spinlock_unlock(&kni_memzone_pool.mutex); > +} > + > + > +/* Shall be called before any allocation happens */ void > +rte_kni_init(unsigned int max_kni_ifaces) { > + unsigned i; > + struct rte_kni_memzone_slot* it; > + const struct rte_memzone *mz; > +#define OBJNAMSIZ 32 > + char obj_name[OBJNAMSIZ]; > + char mz_name[RTE_MEMZONE_NAMESIZE]; > + > + if(max_kni_ifaces =3D=3D 0) { > + //Panic > + RTE_LOG(ERR, KNI, "Invalid number of max_kni_ifaces %d\n", > + max_kni_ifaces); > + rte_panic("Unable to initialize KNI\n"); > + } > + > + //Allocate slot objects > + kni_memzone_pool.slots =3D (struct > rte_kni_memzone_slot*)rte_malloc(NULL, > + sizeof(struct rte_kni_memzone_slot) * > + max_kni_ifaces, > + 0); > + KNI_MEM_CHECK(kni_memzone_pool.slots =3D=3D NULL); > + > + //Initialize general stuff > + kni_memzone_pool.initialized =3D 1; > + kni_memzone_pool.max_ifaces =3D max_kni_ifaces; > + kni_memzone_pool.free =3D &kni_memzone_pool.slots[0]; > + > + //Pre-allocate all memzones of the slots; panic on error > + for(i=3D0; i + > + //Recover current slot > + it =3D &kni_memzone_pool.slots[i]; > + it->id =3D i; > + > + //Allocate KNI context > + snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "KNI_INFO_%d", i); > + mz =3D kni_memzone_reserve(mz_name, sizeof(struct rte_kni), > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_ctx =3D mz; > + > + /* TX RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_tx_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_tx_q =3D mz; > + > + /* RX RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_rx_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_rx_q =3D mz; > + > + /* ALLOC RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_alloc_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_alloc_q =3D mz; > + > + /* FREE RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_free_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_free_q =3D mz; > + > + /* Request RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_req_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_req_q =3D mz; > + > + /* Response RING */ > + snprintf(obj_name, OBJNAMSIZ, "kni_resp_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_resp_q =3D mz; > + > + /* Req/Resp sync mem area */ > + snprintf(obj_name, OBJNAMSIZ, "kni_sync_%d", i); > + mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, > + SOCKET_ID_ANY, 0); > + KNI_MEM_CHECK(mz =3D=3D NULL); > + it->m_sync_addr =3D mz; > + > + if(i+1 =3D=3D max_kni_ifaces) { > + it->next =3D NULL; > + kni_memzone_pool.free_tail =3D it; > + }else > + it->next =3D &kni_memzone_pool.slots[i+1]; > + } > + > + return; > + > +kni_fail: > + rte_panic("Unable to allocate memory for max_kni_ifaces:%d." > + "increase the amount of hugepages memory\n", max_kni_ifaces); } > + > /* It is deprecated and just for backward compatibility */ struct rte_k= ni * > rte_kni_create(uint8_t port_id, @@ -140,14 +328,20 @@ rte_kni_alloc(struc= t > rte_mempool *pktmbuf_pool, > struct rte_kni_device_info dev_info; > struct rte_kni *ctx; > char intf_name[RTE_KNI_NAMESIZE]; > -#define OBJNAMSIZ 32 > - char obj_name[OBJNAMSIZ]; > char mz_name[RTE_MEMZONE_NAMESIZE]; > const struct rte_memzone *mz; > + struct rte_kni_memzone_slot* slot=3DNULL; >=20 > if (!pktmbuf_pool || !conf || !conf->name[0]) > return NULL; >=20 > + /* Check inited */ > + if (kni_memzone_pool.initialized !=3D 1) { > + RTE_LOG(ERR, KNI, "KNI subsystem has not been initialized. " > + "Invoke rte_kni_init() first\n"); > + return NULL; > + } > + > /* Check FD and open once */ > if (kni_fd < 0) { > kni_fd =3D open("/dev/" KNI_DEVICE, O_RDWR); @@ -158,16 +352,23 > @@ rte_kni_alloc(struct rte_mempool *pktmbuf_pool, > } > } >=20 > + /* Get an available slot from the pool */ > + slot =3D kni_memzone_pool_alloc(); > + if(!slot) { > + RTE_LOG(ERR, KNI, "Cannot allocate more KNI interfaces; " > + "increase the number of max_kni_ifaces(current %d) or " > + "deallocate unusued ones.\n", > + kni_memzone_pool.max_ifaces); > + return NULL; > + } > + > + //Recover ctx > + ctx =3D slot->m_ctx->addr; > snprintf(intf_name, RTE_KNI_NAMESIZE, "%s", conf->name); > - snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "KNI_INFO_%s", > intf_name); > - mz =3D kni_memzone_reserve(mz_name, sizeof(struct rte_kni), > - SOCKET_ID_ANY, 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > - ctx =3D mz->addr; >=20 > if (ctx->in_use) { > RTE_LOG(ERR, KNI, "KNI %s is in use\n", ctx->name); > - goto fail; > + return NULL; > } > memset(ctx, 0, sizeof(struct rte_kni)); > if (ops) > @@ -190,83 +391,72 @@ rte_kni_alloc(struct rte_mempool *pktmbuf_pool, > RTE_LOG(INFO, KNI, "pci: %02x:%02x:%02x \t %02x:%02x\n", > dev_info.bus, dev_info.devid, dev_info.function, > dev_info.vendor_id, dev_info.device_id); > - > /* TX RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_tx_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_tx_q; > ctx->tx_q =3D mz->addr; > kni_fifo_init(ctx->tx_q, KNI_FIFO_COUNT_MAX); > dev_info.tx_phys =3D mz->phys_addr; >=20 > /* RX RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_rx_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_rx_q; > ctx->rx_q =3D mz->addr; > kni_fifo_init(ctx->rx_q, KNI_FIFO_COUNT_MAX); > dev_info.rx_phys =3D mz->phys_addr; >=20 > /* ALLOC RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_alloc_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_alloc_q; > ctx->alloc_q =3D mz->addr; > kni_fifo_init(ctx->alloc_q, KNI_FIFO_COUNT_MAX); > dev_info.alloc_phys =3D mz->phys_addr; >=20 > /* FREE RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_free_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_free_q; > ctx->free_q =3D mz->addr; > kni_fifo_init(ctx->free_q, KNI_FIFO_COUNT_MAX); > dev_info.free_phys =3D mz->phys_addr; >=20 > /* Request RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_req_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_req_q; > ctx->req_q =3D mz->addr; > kni_fifo_init(ctx->req_q, KNI_FIFO_COUNT_MAX); > dev_info.req_phys =3D mz->phys_addr; >=20 > /* Response RING */ > - snprintf(obj_name, OBJNAMSIZ, "kni_resp_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_resp_q; > ctx->resp_q =3D mz->addr; > kni_fifo_init(ctx->resp_q, KNI_FIFO_COUNT_MAX); > dev_info.resp_phys =3D mz->phys_addr; >=20 > /* Req/Resp sync mem area */ > - snprintf(obj_name, OBJNAMSIZ, "kni_sync_%s", intf_name); > - mz =3D kni_memzone_reserve(obj_name, KNI_FIFO_SIZE, SOCKET_ID_ANY, > 0); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + mz =3D slot->m_sync_addr; > ctx->sync_addr =3D mz->addr; > dev_info.sync_va =3D mz->addr; > dev_info.sync_phys =3D mz->phys_addr; >=20 > + > /* MBUF mempool */ > snprintf(mz_name, sizeof(mz_name), RTE_MEMPOOL_OBJ_NAME, > pktmbuf_pool->name); > mz =3D rte_memzone_lookup(mz_name); > - KNI_MZ_CHECK(mz =3D=3D NULL); > + KNI_MEM_CHECK(mz =3D=3D NULL); > dev_info.mbuf_va =3D mz->addr; > dev_info.mbuf_phys =3D mz->phys_addr; > ctx->pktmbuf_pool =3D pktmbuf_pool; > ctx->group_id =3D conf->group_id; > + ctx->slot_id =3D slot->id; > ctx->mbuf_size =3D conf->mbuf_size; >=20 > ret =3D ioctl(kni_fd, RTE_KNI_IOCTL_CREATE, &dev_info); > - KNI_MZ_CHECK(ret < 0); > + KNI_MEM_CHECK(ret < 0); >=20 > ctx->in_use =3D 1; >=20 > return ctx; >=20 > -fail: > - > +kni_fail: > + if(slot) > + kni_memzone_pool_dealloc(&kni_memzone_pool.slots[slot->id]); > + > return NULL; > } >=20 > @@ -287,6 +477,7 @@ int > rte_kni_release(struct rte_kni *kni) > { > struct rte_kni_device_info dev_info; > + unsigned slot_id; >=20 > if (!kni || !kni->in_use) > return -1; > @@ -302,8 +493,19 @@ rte_kni_release(struct rte_kni *kni) > kni_free_fifo(kni->rx_q); > kni_free_fifo(kni->alloc_q); > kni_free_fifo(kni->free_q); > + > + slot_id =3D kni->slot_id; > + > + //Memset > memset(kni, 0, sizeof(struct rte_kni)); >=20 > + //Release memzone > + if(slot_id > kni_memzone_pool.max_ifaces) { > + rte_panic("KNI pool: corrupted slot ID: %d, max: %d\n", > + slot_id, kni_memzone_pool.max_ifaces); > + } > + kni_memzone_pool_dealloc(&kni_memzone_pool.slots[slot_id]); > + > return 0; > } >=20 > @@ -437,23 +639,21 @@ rte_kni_get_port_id(struct rte_kni *kni) struct > rte_kni * rte_kni_get(const char *name) { > - struct rte_kni *kni; > - const struct rte_memzone *mz; > - char mz_name[RTE_MEMZONE_NAMESIZE]; > - > - if (!name || !name[0]) > - return NULL; > - > - snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "KNI_INFO_%s", name); > - mz =3D rte_memzone_lookup(mz_name); > - if (!mz) > - return NULL; > - > - kni =3D mz->addr; > - if (!kni->in_use) > - return NULL; > + unsigned i; > + struct rte_kni_memzone_slot* it; > + struct rte_kni* kni; > + > + //Could be improved perf-wise if necessary > + for(i=3D0; i + it =3D &kni_memzone_pool.slots[i]; > + if(it->in_use =3D=3D 0) > + continue; > + kni =3D it->m_ctx->addr; > + if(strncmp(kni->name, name, RTE_KNI_NAMESIZE) =3D=3D 0) > + return kni; > + } >=20 > - return kni; > + return NULL; > } >=20 > /* > diff --git a/lib/librte_kni/rte_kni.h b/lib/librte_kni/rte_kni.h index > 1a0b004..0159a1d 100644 > --- a/lib/librte_kni/rte_kni.h > +++ b/lib/librte_kni/rte_kni.h > @@ -90,11 +90,27 @@ struct rte_kni_conf { }; >=20 > /** > + * Initialize and preallocate KNI subsystem > + * > + * This function is to be executed on the MASTER lcore only, after EAL > + * initialization and before any KNI interface is attempted to be > + * allocated > + * > + * @param max_kni_ifaces > + * The maximum number of KNI interfaces that can coexist concurrently > +*/ extern void rte_kni_init(unsigned int max_kni_ifaces); > + > + > +/** > * Allocate KNI interface according to the port id, mbuf size, mbuf pool= , > * configurations and callbacks for kernel requests.The KNI interface cr= eated > * in the kernel space is the net interface the traditional Linux applic= ation > * talking to. > * > + * The rte_kni_alloc shall not be called before rte_kni_init() has been > + * called. rte_kni_alloc is thread safe. > + * > * @param pktmbuf_pool > * The mempool for allocting mbufs for packets. > * @param conf > @@ -138,6 +154,8 @@ extern struct rte_kni *rte_kni_create(uint8_t port_id= , > * Release KNI interface according to the context. It will also release = the > * paired KNI interface in kernel space. All processing on the specific = KNI > * context need to be stopped before calling this interface. > + * > + * rte_kni_release is thread safe. > * > * @param kni > * The pointer to the context of an existent KNI interface. > -- > 1.7.10.4 Regards, Helin