From: Maxime Coquelin <maxime.coquelin@redhat.com>
To: Nic Chautru <nicolas.chautru@intel.com>,
dev@dpdk.org, thomas@monjalon.net
Cc: trix@redhat.com, mdr@ashroe.eu, bruce.richardson@intel.com,
hemant.agrawal@nxp.com, david.marchand@redhat.com,
stephen@networkplumber.org, hernan.vargas@intel.com
Subject: Re: [PATCH v4 11/14] baseband/acc: support interrupt
Date: Fri, 23 Sep 2022 11:17:06 +0200 [thread overview]
Message-ID: <5c2b9cf5-c600-32f5-7567-b9bb24f79ea0@redhat.com> (raw)
In-Reply-To: <1663806460-45162-12-git-send-email-nicolas.chautru@intel.com>
On 9/22/22 02:27, Nic Chautru wrote:
> Adding support for capability and functions for
> MSI/MSI-X interrupt and underlying information ring.
>
> Signed-off-by: Nic Chautru <nicolas.chautru@intel.com>
> ---
> drivers/baseband/acc/rte_acc200_pmd.c | 346 +++++++++++++++++++++++++++++++++-
> 1 file changed, 344 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/baseband/acc/rte_acc200_pmd.c b/drivers/baseband/acc/rte_acc200_pmd.c
> index fa23b6e..b8b621a 100644
> --- a/drivers/baseband/acc/rte_acc200_pmd.c
> +++ b/drivers/baseband/acc/rte_acc200_pmd.c
> @@ -220,6 +220,193 @@
> acc_conf->q_fft.aq_depth_log2);
> }
>
> +/* Checks PF Info Ring to find the interrupt cause and handles it accordingly */
> +static inline void
> +acc200_check_ir(struct acc_device *acc200_dev)
> +{
> + volatile union acc_info_ring_data *ring_data;
> + uint16_t info_ring_head = acc200_dev->info_ring_head;
> + if (acc200_dev->info_ring == NULL)
> + return;
> +
> + ring_data = acc200_dev->info_ring + (acc200_dev->info_ring_head &
> + ACC_INFO_RING_MASK);
> +
> + while (ring_data->valid) {
> + if ((ring_data->int_nb < ACC200_PF_INT_DMA_DL_DESC_IRQ) || (
> + ring_data->int_nb >
> + ACC200_PF_INT_DMA_DL5G_DESC_IRQ)) {
> + rte_bbdev_log(WARNING, "InfoRing: ITR:%d Info:0x%x",
> + ring_data->int_nb, ring_data->detailed_info);
> + /* Initialize Info Ring entry and move forward */
> + ring_data->val = 0;
> + }
> + info_ring_head++;
> + ring_data = acc200_dev->info_ring +
> + (info_ring_head & ACC_INFO_RING_MASK);
> + }
> +}
> +
> +/* Checks PF Info Ring to find the interrupt cause and handles it accordingly */
> +static inline void
> +acc200_pf_interrupt_handler(struct rte_bbdev *dev)
> +{
> + struct acc_device *acc200_dev = dev->data->dev_private;
> + volatile union acc_info_ring_data *ring_data;
> + struct acc_deq_intr_details deq_intr_det;
> +
> + ring_data = acc200_dev->info_ring + (acc200_dev->info_ring_head &
> + ACC_INFO_RING_MASK);
> +
> + while (ring_data->valid) {
> +
> + rte_bbdev_log_debug(
> + "ACC200 PF Interrupt received, Info Ring data: 0x%x -> %d",
> + ring_data->val, ring_data->int_nb);
> +
> + switch (ring_data->int_nb) {
> + case ACC200_PF_INT_DMA_DL_DESC_IRQ:
> + case ACC200_PF_INT_DMA_UL_DESC_IRQ:
> + case ACC200_PF_INT_DMA_FFT_DESC_IRQ:
> + case ACC200_PF_INT_DMA_UL5G_DESC_IRQ:
> + case ACC200_PF_INT_DMA_DL5G_DESC_IRQ:
> + deq_intr_det.queue_id = get_queue_id_from_ring_info(
> + dev->data, *ring_data);
> + if (deq_intr_det.queue_id == UINT16_MAX) {
> + rte_bbdev_log(ERR,
> + "Couldn't find queue: aq_id: %u, qg_id: %u, vf_id: %u",
> + ring_data->aq_id,
> + ring_data->qg_id,
> + ring_data->vf_id);
> + return;
> + }
> + rte_bbdev_pmd_callback_process(dev,
> + RTE_BBDEV_EVENT_DEQUEUE, &deq_intr_det);
> + break;
> + default:
> + rte_bbdev_pmd_callback_process(dev,
> + RTE_BBDEV_EVENT_ERROR, NULL);
> + break;
> + }
> +
> + /* Initialize Info Ring entry and move forward */
> + ring_data->val = 0;
> + ++acc200_dev->info_ring_head;
> + ring_data = acc200_dev->info_ring +
> + (acc200_dev->info_ring_head & ACC_INFO_RING_MASK);
> + }
> +}
> +
> +/* Checks VF Info Ring to find the interrupt cause and handles it accordingly */
> +static inline void
> +acc200_vf_interrupt_handler(struct rte_bbdev *dev)
> +{
> + struct acc_device *acc200_dev = dev->data->dev_private;
> + volatile union acc_info_ring_data *ring_data;
> + struct acc_deq_intr_details deq_intr_det;
> +
> + ring_data = acc200_dev->info_ring + (acc200_dev->info_ring_head & ACC_INFO_RING_MASK);
> +
> + while (ring_data->valid) {
> +
> + rte_bbdev_log_debug(
> + "ACC200 VF Interrupt received, Info Ring data: 0x%x\n",
> + ring_data->val);
> +
> + switch (ring_data->int_nb) {
> + case ACC200_VF_INT_DMA_DL_DESC_IRQ:
> + case ACC200_VF_INT_DMA_UL_DESC_IRQ:
> + case ACC200_VF_INT_DMA_FFT_DESC_IRQ:
> + case ACC200_VF_INT_DMA_UL5G_DESC_IRQ:
> + case ACC200_VF_INT_DMA_DL5G_DESC_IRQ:
> + /* VFs are not aware of their vf_id - it's set to 0 in
> + * queue structures.
> + */
> + ring_data->vf_id = 0;
> + deq_intr_det.queue_id = get_queue_id_from_ring_info(
> + dev->data, *ring_data);
> + if (deq_intr_det.queue_id == UINT16_MAX) {
> + rte_bbdev_log(ERR,
> + "Couldn't find queue: aq_id: %u, qg_id: %u",
> + ring_data->aq_id,
> + ring_data->qg_id);
> + return;
> + }
> + rte_bbdev_pmd_callback_process(dev,
> + RTE_BBDEV_EVENT_DEQUEUE, &deq_intr_det);
> + break;
> + default:
> + rte_bbdev_pmd_callback_process(dev,
> + RTE_BBDEV_EVENT_ERROR, NULL);
> + break;
> + }
> +
> + /* Initialize Info Ring entry and move forward */
> + ring_data->valid = 0;
> + ++acc200_dev->info_ring_head;
> + ring_data = acc200_dev->info_ring + (acc200_dev->info_ring_head
> + & ACC_INFO_RING_MASK);
> + }
> +}
> +
> +/* Interrupt handler triggered by ACC200 dev for handling specific interrupt */
> +static void
> +acc200_dev_interrupt_handler(void *cb_arg)
> +{
> + struct rte_bbdev *dev = cb_arg;
> + struct acc_device *acc200_dev = dev->data->dev_private;
> +
> + /* Read info ring */
> + if (acc200_dev->pf_device)
> + acc200_pf_interrupt_handler(dev);
> + else
> + acc200_vf_interrupt_handler(dev);
> +}
Both handlers are quite identical, it should be possible to rework
differently so that more code are shared between PF and VF cases.
> +
> +/* Allocate and setup inforing */
> +static int
> +allocate_info_ring(struct rte_bbdev *dev)
> +{
> + struct acc_device *d = dev->data->dev_private;
> + const struct acc200_registry_addr *reg_addr;
> + rte_iova_t info_ring_iova;
> + uint32_t phys_low, phys_high;
> +
> + if (d->info_ring != NULL)
> + return 0; /* Already configured */
> +
> + /* Choose correct registry addresses for the device type */
> + if (d->pf_device)
> + reg_addr = &pf_reg_addr;
> + else
> + reg_addr = &vf_reg_addr;
> + /* Allocate InfoRing */
> + if (d->info_ring == NULL)
This check is useless since it is already done earlier.
> + d->info_ring = rte_zmalloc_socket("Info Ring",
> + ACC_INFO_RING_NUM_ENTRIES *
> + sizeof(*d->info_ring), RTE_CACHE_LINE_SIZE,
> + dev->data->socket_id);
> + if (d->info_ring == NULL) {
> + rte_bbdev_log(ERR,
> + "Failed to allocate Info Ring for %s:%u",
> + dev->device->driver->name,
> + dev->data->dev_id);
> + return -ENOMEM;
> + }
> + info_ring_iova = rte_malloc_virt2iova(d->info_ring);
> +
> + /* Setup Info Ring */
> + phys_high = (uint32_t)(info_ring_iova >> 32);
> + phys_low = (uint32_t)(info_ring_iova);
> + acc_reg_write(d, reg_addr->info_ring_hi, phys_high);
> + acc_reg_write(d, reg_addr->info_ring_lo, phys_low);
> + acc_reg_write(d, reg_addr->info_ring_en, ACC200_REG_IRQ_EN_ALL);
> + d->info_ring_head = (acc_reg_read(d, reg_addr->info_ring_ptr) &
> + 0xFFF) / sizeof(union acc_info_ring_data);
> + return 0;
> +}
> +
> +
> /* Allocate 64MB memory used for all software rings */
> static int
> acc200_setup_queues(struct rte_bbdev *dev, uint16_t num_queues, int socket_id)
> @@ -227,6 +414,7 @@
> uint32_t phys_low, phys_high, value;
> struct acc_device *d = dev->data->dev_private;
> const struct acc200_registry_addr *reg_addr;
> + int ret;
>
> if (d->pf_device && !d->acc_conf.pf_mode_en) {
> rte_bbdev_log(NOTICE,
> @@ -327,6 +515,14 @@
> acc_reg_write(d, reg_addr->tail_ptrs_fft_hi, phys_high);
> acc_reg_write(d, reg_addr->tail_ptrs_fft_lo, phys_low);
>
> + ret = allocate_info_ring(dev);
> + if (ret < 0) {
> + rte_bbdev_log(ERR, "Failed to allocate info_ring for %s:%u",
> + dev->device->driver->name,
> + dev->data->dev_id);
> + /* Continue */
> + }
> +
> if (d->harq_layout == NULL)
> d->harq_layout = rte_zmalloc_socket("HARQ Layout",
> ACC_HARQ_LAYOUT * sizeof(*d->harq_layout),
> @@ -349,17 +545,121 @@
> return 0;
> }
>
> +static int
> +acc200_intr_enable(struct rte_bbdev *dev)
> +{
> + int ret;
> + struct acc_device *d = dev->data->dev_private;
> + /*
> + * MSI/MSI-X are supported
> + * Option controlled by vfio-intr through EAL parameter
> + */
> + if (rte_intr_type_get(dev->intr_handle) == RTE_INTR_HANDLE_VFIO_MSI) {
> +
> + ret = allocate_info_ring(dev);
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't allocate info ring for device: %s",
> + dev->data->name);
> + return ret;
> + }
> + ret = rte_intr_enable(dev->intr_handle);
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't enable interrupts for device: %s",
> + dev->data->name);
> + rte_free(d->info_ring);
> + return ret;
> + }
> + ret = rte_intr_callback_register(dev->intr_handle,
> + acc200_dev_interrupt_handler, dev);
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't register interrupt callback for device: %s",
> + dev->data->name);
> + rte_free(d->info_ring);
> + return ret;
> + }
> +
> + return 0;
> + } else if (rte_intr_type_get(dev->intr_handle) == RTE_INTR_HANDLE_VFIO_MSIX) {
> +
> + ret = allocate_info_ring(dev);
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't allocate info ring for device: %s",
> + dev->data->name);
> + return ret;
> + }
> +
> + int i, max_queues;
> + struct acc_device *acc200_dev = dev->data->dev_private;
Don't mix declarations & code.
> + if (acc200_dev->pf_device)
> + max_queues = ACC200_MAX_PF_MSIX;
> + else
> + max_queues = ACC200_MAX_VF_MSIX;
> +
> + if (rte_intr_efd_enable(dev->intr_handle, max_queues)) {
> + rte_bbdev_log(ERR, "Failed to create fds for %u queues",
> + dev->data->num_queues);
> + return -1;
> + }
> +
> + for (i = 0; i < max_queues; ++i) {
> + if (rte_intr_efds_index_set(dev->intr_handle, i,
> + rte_intr_fd_get(dev->intr_handle)))
> + return -rte_errno;
> + }
> +
> + if (rte_intr_vec_list_alloc(dev->intr_handle, "intr_vec",
> + dev->data->num_queues)) {
> + rte_bbdev_log(ERR, "Failed to allocate %u vectors",
> + dev->data->num_queues);
> + return -ENOMEM;
> + }
> +
> + ret = rte_intr_enable(dev->intr_handle);
> +
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't enable interrupts for device: %s",
> + dev->data->name);
> + rte_free(d->info_ring);
> + return ret;
> + }
> + ret = rte_intr_callback_register(dev->intr_handle,
> + acc200_dev_interrupt_handler, dev);
> + if (ret < 0) {
> + rte_bbdev_log(ERR,
> + "Couldn't register interrupt callback for device: %s",
> + dev->data->name);
> + rte_free(d->info_ring);
> + return ret;
> + }
> +
> + return 0;
> + }
> +
> + rte_bbdev_log(ERR, "ACC200 (%s) supports only VFIO MSI/MSI-X interrupts\n",
> + dev->data->name);
> + return -ENOTSUP;
> +}
> +
> /* Free memory used for software rings */
> static int
> acc200_dev_close(struct rte_bbdev *dev)
> {
> struct acc_device *d = dev->data->dev_private;
> + acc200_check_ir(d);
> if (d->sw_rings_base != NULL) {
> rte_free(d->tail_ptrs);
> + rte_free(d->info_ring);
> rte_free(d->sw_rings_base);
> rte_free(d->harq_layout);
> d->sw_rings_base = NULL;
> d->tail_ptrs = NULL;
> + d->info_ring = NULL;
> d->harq_layout = NULL;
> }
> /* Ensure all in flight HW transactions are completed */
> @@ -652,6 +952,7 @@
> RTE_BBDEV_TURBO_CONTINUE_CRC_MATCH |
> RTE_BBDEV_TURBO_SOFT_OUTPUT |
> RTE_BBDEV_TURBO_EARLY_TERMINATION |
> + RTE_BBDEV_TURBO_DEC_INTERRUPTS |
> RTE_BBDEV_TURBO_NEG_LLR_1_BIT_IN |
> RTE_BBDEV_TURBO_NEG_LLR_1_BIT_SOFT_OUT |
> RTE_BBDEV_TURBO_MAP_DEC |
> @@ -673,6 +974,7 @@
> RTE_BBDEV_TURBO_CRC_24B_ATTACH |
> RTE_BBDEV_TURBO_RV_INDEX_BYPASS |
> RTE_BBDEV_TURBO_RATE_MATCH |
> + RTE_BBDEV_TURBO_ENC_INTERRUPTS |
> RTE_BBDEV_TURBO_ENC_SCATTER_GATHER,
> .num_buffers_src =
> RTE_BBDEV_TURBO_MAX_CODE_BLOCKS,
> @@ -686,7 +988,8 @@
> .capability_flags =
> RTE_BBDEV_LDPC_RATE_MATCH |
> RTE_BBDEV_LDPC_CRC_24B_ATTACH |
> - RTE_BBDEV_LDPC_INTERLEAVER_BYPASS,
> + RTE_BBDEV_LDPC_INTERLEAVER_BYPASS |
> + RTE_BBDEV_LDPC_ENC_INTERRUPTS,
> .num_buffers_src =
> RTE_BBDEV_LDPC_MAX_CODE_BLOCKS,
> .num_buffers_dst =
> @@ -707,7 +1010,8 @@
> RTE_BBDEV_LDPC_DEINTERLEAVER_BYPASS |
> RTE_BBDEV_LDPC_DEC_SCATTER_GATHER |
> RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION |
> - RTE_BBDEV_LDPC_LLR_COMPRESSION,
> + RTE_BBDEV_LDPC_LLR_COMPRESSION |
> + RTE_BBDEV_LDPC_DEC_INTERRUPTS,
> .llr_size = 8,
> .llr_decimals = 1,
> .num_buffers_src =
> @@ -775,15 +1079,46 @@
> dev_info->min_alignment = 1;
> dev_info->capabilities = bbdev_capabilities;
> dev_info->harq_buffer_size = 0;
> +
> + acc200_check_ir(d);
> +}
> +
> +static int
> +acc200_queue_intr_enable(struct rte_bbdev *dev, uint16_t queue_id)
> +{
> + struct acc_queue *q = dev->data->queues[queue_id].queue_private;
> +
> + if (rte_intr_type_get(dev->intr_handle) != RTE_INTR_HANDLE_VFIO_MSI &&
> + rte_intr_type_get(dev->intr_handle) != RTE_INTR_HANDLE_VFIO_MSIX)
> + return -ENOTSUP;
> +
> + q->irq_enable = 1;
> + return 0;
> +}
> +
> +static int
> +acc200_queue_intr_disable(struct rte_bbdev *dev, uint16_t queue_id)
> +{
> + struct acc_queue *q = dev->data->queues[queue_id].queue_private;
> +
> + if (rte_intr_type_get(dev->intr_handle) != RTE_INTR_HANDLE_VFIO_MSI &&
> + rte_intr_type_get(dev->intr_handle) != RTE_INTR_HANDLE_VFIO_MSIX)
> + return -ENOTSUP;
> +
> + q->irq_enable = 0;
> + return 0;
> }
>
> static const struct rte_bbdev_ops acc200_bbdev_ops = {
> .setup_queues = acc200_setup_queues,
> + .intr_enable = acc200_intr_enable,
> .close = acc200_dev_close,
> .info_get = acc200_dev_info_get,
> .queue_setup = acc200_queue_setup,
> .queue_release = acc200_queue_release,
> .queue_stop = acc_queue_stop,
> + .queue_intr_enable = acc200_queue_intr_enable,
> + .queue_intr_disable = acc200_queue_intr_disable
> };
>
> /* ACC200 PCI PF address map */
> @@ -2694,6 +3029,7 @@
> if (op->status != 0) {
> /* These errors are not expected */
> q_data->queue_stats.dequeue_err_count++;
> + acc200_check_ir(q->d);
> }
>
> /* CRC invalid if error exists */
> @@ -2763,6 +3099,9 @@
>
> op->ldpc_dec.iter_count = (uint8_t) rsp.iter_cnt;
>
> + if (op->status & (1 << RTE_BBDEV_DRV_ERROR))
> + acc200_check_ir(q->d);
> +
> /* Check if this is the last desc in batch (Atomic Queue) */
> if (desc->req.last_desc_in_batch) {
> (*aq_dequeued)++;
> @@ -3232,6 +3571,9 @@
> if (op->status != 0)
> q_data->queue_stats.dequeue_err_count++;
>
> + if (op->status & (1 << RTE_BBDEV_DRV_ERROR))
> + acc200_check_ir(q->d);
> +
> /* Check if this is the last desc in batch (Atomic Queue) */
> if (desc->req.last_desc_in_batch) {
> (*aq_dequeued)++;
next prev parent reply other threads:[~2022-09-23 9:17 UTC|newest]
Thread overview: 43+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-22 0:27 From: Nic Chautru <nicolas.chautru@intel.com> Nic Chautru
2022-09-22 0:27 ` [PATCH v4 01/14] baseband/acc100: remove unused registers Nic Chautru
2022-09-22 0:37 ` Chautru, Nicolas
2022-09-22 13:08 ` Maxime Coquelin
2022-09-22 13:09 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 02/14] baseband/acc100: refactor to segregate common code Nic Chautru
2022-09-22 13:15 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 03/14] baseband/acc: rename directory from acc100 to acc Nic Chautru
2022-09-22 13:17 ` Maxime Coquelin
2022-09-22 18:06 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 04/14] baseband/acc: introduce PMD for ACC200 Nic Chautru
2022-09-22 14:01 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 05/14] baseband/acc: add HW register definitions " Nic Chautru
2022-09-22 14:04 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 06/14] baseband/acc: add info get function " Nic Chautru
2022-09-22 14:11 ` Maxime Coquelin
2022-09-22 18:20 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 07/14] baseband/acc: add queue configuration " Nic Chautru
2022-09-22 14:30 ` Maxime Coquelin
2022-09-22 18:51 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 08/14] baseband/acc: add LDPC processing functions Nic Chautru
2022-09-23 8:29 ` Maxime Coquelin
2022-09-23 21:55 ` Chautru, Nicolas
2022-09-27 13:12 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 09/14] baseband/acc: add LTE " Nic Chautru
2022-09-23 8:59 ` Maxime Coquelin
2022-09-23 22:21 ` Chautru, Nicolas
2022-09-27 13:33 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 10/14] baseband/acc: add support for FFT operations Nic Chautru
2022-09-23 9:05 ` Maxime Coquelin
2022-09-23 22:31 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 11/14] baseband/acc: support interrupt Nic Chautru
2022-09-23 9:17 ` Maxime Coquelin [this message]
2022-09-23 22:58 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 12/14] baseband/acc: add device status and vf2pf comms Nic Chautru
2022-09-23 9:23 ` Maxime Coquelin
2022-09-24 0:04 ` Chautru, Nicolas
2022-09-22 0:27 ` [PATCH v4 13/14] baseband/acc: add PF configure companion function Nic Chautru
2022-09-23 9:26 ` Maxime Coquelin
2022-09-24 0:20 ` Chautru, Nicolas
2022-09-27 13:56 ` Maxime Coquelin
2022-09-22 0:27 ` [PATCH v4 14/14] baseband/acc: simplify meson dependency Nic Chautru
2022-09-23 11:41 ` From: Nic Chautru <nicolas.chautru@intel.com> Maxime Coquelin
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=5c2b9cf5-c600-32f5-7567-b9bb24f79ea0@redhat.com \
--to=maxime.coquelin@redhat.com \
--cc=bruce.richardson@intel.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=hemant.agrawal@nxp.com \
--cc=hernan.vargas@intel.com \
--cc=mdr@ashroe.eu \
--cc=nicolas.chautru@intel.com \
--cc=stephen@networkplumber.org \
--cc=thomas@monjalon.net \
--cc=trix@redhat.com \
/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).