From 4639d60d2bfb7f5007b5d93788fd93c19b63f000 Mon Sep 17 00:00:00 2001 From: Tomer Tayar Date: Mon, 7 Dec 2015 06:25:56 -0500 Subject: [PATCH 1/4] qed: Fix corner case for chain in-between pages The amount of chain next pointer elements between the producer and the consumer indices depends on which pages they currently point to. The current calculation is based only on their difference, and it can lead to a number of free elements which is higher by 1 than the actual value. Signed-off-by: Tomer Tayar Signed-off-by: Manish Chopra Signed-off-by: David S. Miller --- include/linux/qed/qed_chain.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/include/linux/qed/qed_chain.h b/include/linux/qed/qed_chain.h index b920c3605c46..41b9049b57e2 100644 --- a/include/linux/qed/qed_chain.h +++ b/include/linux/qed/qed_chain.h @@ -111,7 +111,8 @@ static inline u16 qed_chain_get_elem_left(struct qed_chain *p_chain) used = ((u32)0x10000u + (u32)(p_chain->prod_idx)) - (u32)p_chain->cons_idx; if (p_chain->mode == QED_CHAIN_MODE_NEXT_PTR) - used -= (used / p_chain->elem_per_page); + used -= p_chain->prod_idx / p_chain->elem_per_page - + p_chain->cons_idx / p_chain->elem_per_page; return p_chain->capacity - used; } From 76a9a3642a0b72d5687d680150580d55b6ea9804 Mon Sep 17 00:00:00 2001 From: Tomer Tayar Date: Mon, 7 Dec 2015 06:25:57 -0500 Subject: [PATCH 2/4] qed: fix handling of concurrent ramrods. Concurrent non-blocking slowpath ramrods can be completed out-of-order on the completion chain. Recycling completed elements, while previously sent elements are still completion pending, can lead to overriding of active elements on the chain. Furthermore, sending pending slowpath ramrods currently lacks the update of the chain element physical pointer. This patch: * Ensures that ramrods are sent to the FW with consecutive echo values. * Handles out-of-order completions by freeing only first successive completed entries. * Updates the chain element physical pointer when copying a pending element into a free element for sending. Signed-off-by: Tomer Tayar Signed-off-by: Manish Chopra Signed-off-by: David S. Miller --- drivers/net/ethernet/qlogic/qed/qed_sp.h | 8 ++- drivers/net/ethernet/qlogic/qed/qed_spq.c | 69 +++++++++++++++++------ include/linux/qed/common_hsi.h | 2 + 3 files changed, 59 insertions(+), 20 deletions(-) diff --git a/drivers/net/ethernet/qlogic/qed/qed_sp.h b/drivers/net/ethernet/qlogic/qed/qed_sp.h index 31a1f1eb4f56..287fadfab52d 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_sp.h +++ b/drivers/net/ethernet/qlogic/qed/qed_sp.h @@ -124,8 +124,12 @@ struct qed_spq { dma_addr_t p_phys; struct qed_spq_entry *p_virt; - /* Used as index for completions (returns on EQ by FW) */ - u16 echo_idx; +#define SPQ_RING_SIZE \ + (CORE_SPQE_PAGE_SIZE_BYTES / sizeof(struct slow_path_element)) + + /* Bitmap for handling out-of-order completions */ + DECLARE_BITMAP(p_comp_bitmap, SPQ_RING_SIZE); + u8 comp_bitmap_idx; /* Statistics */ u32 unlimited_pending_count; diff --git a/drivers/net/ethernet/qlogic/qed/qed_spq.c b/drivers/net/ethernet/qlogic/qed/qed_spq.c index 7c0b8459666e..3dd548ab8df1 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_spq.c +++ b/drivers/net/ethernet/qlogic/qed/qed_spq.c @@ -112,8 +112,6 @@ static int qed_spq_fill_entry(struct qed_hwfn *p_hwfn, struct qed_spq_entry *p_ent) { - p_ent->elem.hdr.echo = 0; - p_hwfn->p_spq->echo_idx++; p_ent->flags = 0; switch (p_ent->comp_mode) { @@ -195,10 +193,12 @@ static int qed_spq_hw_post(struct qed_hwfn *p_hwfn, struct qed_spq *p_spq, struct qed_spq_entry *p_ent) { - struct qed_chain *p_chain = &p_hwfn->p_spq->chain; + struct qed_chain *p_chain = &p_hwfn->p_spq->chain; + u16 echo = qed_chain_get_prod_idx(p_chain); struct slow_path_element *elem; struct core_db_data db; + p_ent->elem.hdr.echo = cpu_to_le16(echo); elem = qed_chain_produce(p_chain); if (!elem) { DP_NOTICE(p_hwfn, "Failed to produce from SPQ chain\n"); @@ -437,7 +437,9 @@ void qed_spq_setup(struct qed_hwfn *p_hwfn) p_spq->comp_count = 0; p_spq->comp_sent_count = 0; p_spq->unlimited_pending_count = 0; - p_spq->echo_idx = 0; + + bitmap_zero(p_spq->p_comp_bitmap, SPQ_RING_SIZE); + p_spq->comp_bitmap_idx = 0; /* SPQ cid, cannot fail */ qed_cxt_acquire_cid(p_hwfn, PROTOCOLID_CORE, &p_spq->cid); @@ -582,26 +584,32 @@ qed_spq_add_entry(struct qed_hwfn *p_hwfn, struct qed_spq *p_spq = p_hwfn->p_spq; if (p_ent->queue == &p_spq->unlimited_pending) { - struct qed_spq_entry *p_en2; if (list_empty(&p_spq->free_pool)) { list_add_tail(&p_ent->list, &p_spq->unlimited_pending); p_spq->unlimited_pending_count++; return 0; + } else { + struct qed_spq_entry *p_en2; + + p_en2 = list_first_entry(&p_spq->free_pool, + struct qed_spq_entry, + list); + list_del(&p_en2->list); + + /* Copy the ring element physical pointer to the new + * entry, since we are about to override the entire ring + * entry and don't want to lose the pointer. + */ + p_ent->elem.data_ptr = p_en2->elem.data_ptr; + + *p_en2 = *p_ent; + + kfree(p_ent); + + p_ent = p_en2; } - - p_en2 = list_first_entry(&p_spq->free_pool, - struct qed_spq_entry, - list); - list_del(&p_en2->list); - - /* Strcut assignment */ - *p_en2 = *p_ent; - - kfree(p_ent); - - p_ent = p_en2; } /* entry is to be placed in 'pending' queue */ @@ -777,13 +785,38 @@ int qed_spq_completion(struct qed_hwfn *p_hwfn, list_for_each_entry_safe(p_ent, tmp, &p_spq->completion_pending, list) { if (p_ent->elem.hdr.echo == echo) { + u16 pos = le16_to_cpu(echo) % SPQ_RING_SIZE; + list_del(&p_ent->list); - qed_chain_return_produced(&p_spq->chain); + /* Avoid overriding of SPQ entries when getting + * out-of-order completions, by marking the completions + * in a bitmap and increasing the chain consumer only + * for the first successive completed entries. + */ + bitmap_set(p_spq->p_comp_bitmap, pos, SPQ_RING_SIZE); + + while (test_bit(p_spq->comp_bitmap_idx, + p_spq->p_comp_bitmap)) { + bitmap_clear(p_spq->p_comp_bitmap, + p_spq->comp_bitmap_idx, + SPQ_RING_SIZE); + p_spq->comp_bitmap_idx++; + qed_chain_return_produced(&p_spq->chain); + } + p_spq->comp_count++; found = p_ent; break; } + + /* This is relatively uncommon - depends on scenarios + * which have mutliple per-PF sent ramrods. + */ + DP_VERBOSE(p_hwfn, QED_MSG_SPQ, + "Got completion for echo %04x - doesn't match echo %04x in completion pending list\n", + le16_to_cpu(echo), + le16_to_cpu(p_ent->elem.hdr.echo)); } /* Release lock before callback, as callback may post diff --git a/include/linux/qed/common_hsi.h b/include/linux/qed/common_hsi.h index 6a4347639c03..1d1ba2c5ee7a 100644 --- a/include/linux/qed/common_hsi.h +++ b/include/linux/qed/common_hsi.h @@ -9,6 +9,8 @@ #ifndef __COMMON_HSI__ #define __COMMON_HSI__ +#define CORE_SPQE_PAGE_SIZE_BYTES 4096 + #define FW_MAJOR_VERSION 8 #define FW_MINOR_VERSION 4 #define FW_REVISION_VERSION 2 From c78df14ee0f6bc5e8741b4324b600b7277abb13e Mon Sep 17 00:00:00 2001 From: Ariel Elior Date: Mon, 7 Dec 2015 06:25:58 -0500 Subject: [PATCH 3/4] qed: Fix BAR size split for some servers Can't rely on pci config space to discover bar size, as in some environments this returns a wrong, too large value. Instead, rely on device register, which contains the value provided by MFW at preboot. Signed-off-by: Ariel Elior Signed-off-by: Manish Chopra Signed-off-by: David S. Miller --- drivers/net/ethernet/qlogic/qed/qed_dev.c | 53 +++++++++++-------- .../net/ethernet/qlogic/qed/qed_reg_addr.h | 4 ++ 2 files changed, 36 insertions(+), 21 deletions(-) diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c index 803b190ccada..817bbd5476ff 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_dev.c +++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c @@ -1385,52 +1385,63 @@ err0: return rc; } -static u32 qed_hw_bar_size(struct qed_dev *cdev, - u8 bar_id) +static u32 qed_hw_bar_size(struct qed_hwfn *p_hwfn, + u8 bar_id) { - u32 size = pci_resource_len(cdev->pdev, (bar_id > 0) ? 2 : 0); + u32 bar_reg = (bar_id == 0 ? PGLUE_B_REG_PF_BAR0_SIZE + : PGLUE_B_REG_PF_BAR1_SIZE); + u32 val = qed_rd(p_hwfn, p_hwfn->p_main_ptt, bar_reg); - return size / cdev->num_hwfns; + /* Get the BAR size(in KB) from hardware given val */ + return 1 << (val + 15); } int qed_hw_prepare(struct qed_dev *cdev, int personality) { - int rc, i; + struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev); + int rc; /* Store the precompiled init data ptrs */ qed_init_iro_array(cdev); /* Initialize the first hwfn - will learn number of hwfns */ - rc = qed_hw_prepare_single(&cdev->hwfns[0], cdev->regview, + rc = qed_hw_prepare_single(p_hwfn, + cdev->regview, cdev->doorbells, personality); if (rc) return rc; - personality = cdev->hwfns[0].hw_info.personality; + personality = p_hwfn->hw_info.personality; /* Initialize the rest of the hwfns */ - for (i = 1; i < cdev->num_hwfns; i++) { + if (cdev->num_hwfns > 1) { void __iomem *p_regview, *p_doorbell; + u8 __iomem *addr; - p_regview = cdev->regview + - i * qed_hw_bar_size(cdev, 0); - p_doorbell = cdev->doorbells + - i * qed_hw_bar_size(cdev, 1); - rc = qed_hw_prepare_single(&cdev->hwfns[i], p_regview, + /* adjust bar offset for second engine */ + addr = cdev->regview + qed_hw_bar_size(p_hwfn, 0) / 2; + p_regview = addr; + + /* adjust doorbell bar offset for second engine */ + addr = cdev->doorbells + qed_hw_bar_size(p_hwfn, 1) / 2; + p_doorbell = addr; + + /* prepare second hw function */ + rc = qed_hw_prepare_single(&cdev->hwfns[1], p_regview, p_doorbell, personality); + + /* in case of error, need to free the previously + * initiliazed hwfn 0. + */ if (rc) { - /* Cleanup previously initialized hwfns */ - while (--i >= 0) { - qed_init_free(&cdev->hwfns[i]); - qed_mcp_free(&cdev->hwfns[i]); - qed_hw_hwfn_free(&cdev->hwfns[i]); - } - return rc; + qed_init_free(p_hwfn); + qed_mcp_free(p_hwfn); + qed_hw_hwfn_free(p_hwfn); } } - return 0; + return rc; } void qed_hw_remove(struct qed_dev *cdev) diff --git a/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h b/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h index 7a5ce5914ace..e8df12335a97 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h +++ b/drivers/net/ethernet/qlogic/qed/qed_reg_addr.h @@ -363,4 +363,8 @@ 0x7 << 0) #define MCP_REG_NVM_CFG4_FLASH_SIZE_SHIFT \ 0 +#define PGLUE_B_REG_PF_BAR0_SIZE \ + 0x2aae60UL +#define PGLUE_B_REG_PF_BAR1_SIZE \ + 0x2aae64UL #endif From 8f16bc97fa2a47e2e46d36f2f682e1215ee172f5 Mon Sep 17 00:00:00 2001 From: Sudarsana Kalluru Date: Mon, 7 Dec 2015 06:25:59 -0500 Subject: [PATCH 4/4] qed: Correct slowpath interrupt scheme When using INTa, ISR might be called before device is configured for INTa [E.g., due to other device asserting the shared interrupt line], in which case the ISR would read the SISR registers that shouldn't be read unless HW is already configured for INTa. This might break interrupts later on. There's also an MSI-X issue due to this difference, although it's mostly theoretical. This patch changes the initialization order, calling request_irq() for the slowpath interrupt only after the chip is configured for working in the preferred interrupt mode. Signed-off-by: Sudarsana Kalluru Signed-off-by: Manish Chopra Signed-off-by: David S. Miller --- drivers/net/ethernet/qlogic/qed/qed.h | 3 ++ drivers/net/ethernet/qlogic/qed/qed_int.c | 33 +++++++++---- drivers/net/ethernet/qlogic/qed/qed_int.h | 15 ++++-- drivers/net/ethernet/qlogic/qed/qed_main.c | 56 +++++++--------------- 4 files changed, 55 insertions(+), 52 deletions(-) diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h index ac17d8669b1a..1292c360390c 100644 --- a/drivers/net/ethernet/qlogic/qed/qed.h +++ b/drivers/net/ethernet/qlogic/qed/qed.h @@ -299,6 +299,7 @@ struct qed_hwfn { /* Flag indicating whether interrupts are enabled or not*/ bool b_int_enabled; + bool b_int_requested; struct qed_mcp_info *mcp_info; @@ -491,6 +492,8 @@ u32 qed_unzip_data(struct qed_hwfn *p_hwfn, u32 input_len, u8 *input_buf, u32 max_size, u8 *unzip_buf); +int qed_slowpath_irq_req(struct qed_hwfn *hwfn); + #define QED_ETH_INTERFACE_VERSION 300 #endif /* _QED_H */ diff --git a/drivers/net/ethernet/qlogic/qed/qed_int.c b/drivers/net/ethernet/qlogic/qed/qed_int.c index de50e84902af..9cc9d62c1fec 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_int.c +++ b/drivers/net/ethernet/qlogic/qed/qed_int.c @@ -783,22 +783,16 @@ void qed_int_igu_enable_int(struct qed_hwfn *p_hwfn, qed_wr(p_hwfn, p_ptt, IGU_REG_PF_CONFIGURATION, igu_pf_conf); } -void qed_int_igu_enable(struct qed_hwfn *p_hwfn, - struct qed_ptt *p_ptt, - enum qed_int_mode int_mode) +int qed_int_igu_enable(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + enum qed_int_mode int_mode) { - int i; - - p_hwfn->b_int_enabled = 1; + int rc, i; /* Mask non-link attentions */ for (i = 0; i < 9; i++) qed_wr(p_hwfn, p_ptt, MISC_REG_AEU_ENABLE1_IGU_OUT_0 + (i << 2), 0); - /* Enable interrupt Generation */ - qed_int_igu_enable_int(p_hwfn, p_ptt, int_mode); - /* Configure AEU signal change to produce attentions for link */ qed_wr(p_hwfn, p_ptt, IGU_REG_LEADING_EDGE_LATCH, 0xfff); qed_wr(p_hwfn, p_ptt, IGU_REG_TRAILING_EDGE_LATCH, 0xfff); @@ -808,6 +802,19 @@ void qed_int_igu_enable(struct qed_hwfn *p_hwfn, /* Unmask AEU signals toward IGU */ qed_wr(p_hwfn, p_ptt, MISC_REG_AEU_MASK_ATTN_IGU, 0xff); + if ((int_mode != QED_INT_MODE_INTA) || IS_LEAD_HWFN(p_hwfn)) { + rc = qed_slowpath_irq_req(p_hwfn); + if (rc != 0) { + DP_NOTICE(p_hwfn, "Slowpath IRQ request failed\n"); + return -EINVAL; + } + p_hwfn->b_int_requested = true; + } + /* Enable interrupt Generation */ + qed_int_igu_enable_int(p_hwfn, p_ptt, int_mode); + p_hwfn->b_int_enabled = 1; + + return rc; } void qed_int_igu_disable_int(struct qed_hwfn *p_hwfn, @@ -1127,3 +1134,11 @@ int qed_int_get_num_sbs(struct qed_hwfn *p_hwfn, return info->igu_sb_cnt; } + +void qed_int_disable_post_isr_release(struct qed_dev *cdev) +{ + int i; + + for_each_hwfn(cdev, i) + cdev->hwfns[i].b_int_requested = false; +} diff --git a/drivers/net/ethernet/qlogic/qed/qed_int.h b/drivers/net/ethernet/qlogic/qed/qed_int.h index 16b57518e706..51e0b09a7f47 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_int.h +++ b/drivers/net/ethernet/qlogic/qed/qed_int.h @@ -169,10 +169,14 @@ int qed_int_get_num_sbs(struct qed_hwfn *p_hwfn, int *p_iov_blks); /** - * @file + * @brief qed_int_disable_post_isr_release - performs the cleanup post ISR + * release. The API need to be called after releasing all slowpath IRQs + * of the device. + * + * @param cdev * - * @brief Interrupt handler */ +void qed_int_disable_post_isr_release(struct qed_dev *cdev); #define QED_CAU_DEF_RX_TIMER_RES 0 #define QED_CAU_DEF_TX_TIMER_RES 0 @@ -366,10 +370,11 @@ void qed_int_setup(struct qed_hwfn *p_hwfn, * @param p_hwfn * @param p_ptt * @param int_mode + * + * @return int */ -void qed_int_igu_enable(struct qed_hwfn *p_hwfn, - struct qed_ptt *p_ptt, - enum qed_int_mode int_mode); +int qed_int_igu_enable(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt, + enum qed_int_mode int_mode); /** * @brief - Initialize CAU status block entry diff --git a/drivers/net/ethernet/qlogic/qed/qed_main.c b/drivers/net/ethernet/qlogic/qed/qed_main.c index 947c7af72b25..174f7341c5c3 100644 --- a/drivers/net/ethernet/qlogic/qed/qed_main.c +++ b/drivers/net/ethernet/qlogic/qed/qed_main.c @@ -476,41 +476,22 @@ static irqreturn_t qed_single_int(int irq, void *dev_instance) return rc; } -static int qed_slowpath_irq_req(struct qed_dev *cdev) +int qed_slowpath_irq_req(struct qed_hwfn *hwfn) { - int i = 0, rc = 0; + struct qed_dev *cdev = hwfn->cdev; + int rc = 0; + u8 id; if (cdev->int_params.out.int_mode == QED_INT_MODE_MSIX) { - /* Request all the slowpath MSI-X vectors */ - for (i = 0; i < cdev->num_hwfns; i++) { - snprintf(cdev->hwfns[i].name, NAME_SIZE, - "sp-%d-%02x:%02x.%02x", - i, cdev->pdev->bus->number, - PCI_SLOT(cdev->pdev->devfn), - cdev->hwfns[i].abs_pf_id); - - rc = request_irq(cdev->int_params.msix_table[i].vector, - qed_msix_sp_int, 0, - cdev->hwfns[i].name, - cdev->hwfns[i].sp_dpc); - if (rc) - break; - - DP_VERBOSE(&cdev->hwfns[i], - (NETIF_MSG_INTR | QED_MSG_SP), + id = hwfn->my_id; + snprintf(hwfn->name, NAME_SIZE, "sp-%d-%02x:%02x.%02x", + id, cdev->pdev->bus->number, + PCI_SLOT(cdev->pdev->devfn), hwfn->abs_pf_id); + rc = request_irq(cdev->int_params.msix_table[id].vector, + qed_msix_sp_int, 0, hwfn->name, hwfn->sp_dpc); + if (!rc) + DP_VERBOSE(hwfn, (NETIF_MSG_INTR | QED_MSG_SP), "Requested slowpath MSI-X\n"); - } - - if (i != cdev->num_hwfns) { - /* Free already request MSI-X vectors */ - for (i--; i >= 0; i--) { - unsigned int vec = - cdev->int_params.msix_table[i].vector; - synchronize_irq(vec); - free_irq(cdev->int_params.msix_table[i].vector, - cdev->hwfns[i].sp_dpc); - } - } } else { unsigned long flags = 0; @@ -534,13 +515,17 @@ static void qed_slowpath_irq_free(struct qed_dev *cdev) if (cdev->int_params.out.int_mode == QED_INT_MODE_MSIX) { for_each_hwfn(cdev, i) { + if (!cdev->hwfns[i].b_int_requested) + break; synchronize_irq(cdev->int_params.msix_table[i].vector); free_irq(cdev->int_params.msix_table[i].vector, cdev->hwfns[i].sp_dpc); } } else { - free_irq(cdev->pdev->irq, cdev); + if (QED_LEADING_HWFN(cdev)->b_int_requested) + free_irq(cdev->pdev->irq, cdev); } + qed_int_disable_post_isr_release(cdev); } static int qed_nic_stop(struct qed_dev *cdev) @@ -765,16 +750,11 @@ static int qed_slowpath_start(struct qed_dev *cdev, if (rc) goto err1; - /* Request the slowpath IRQ */ - rc = qed_slowpath_irq_req(cdev); - if (rc) - goto err2; - /* Allocate stream for unzipping */ rc = qed_alloc_stream_mem(cdev); if (rc) { DP_NOTICE(cdev, "Failed to allocate stream memory\n"); - goto err3; + goto err2; } /* Start the slowpath */