From patchwork Mon May 4 17:54:19 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Kirti Wankhede X-Patchwork-Id: 11527395 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id D63731668 for ; Mon, 4 May 2020 18:27:45 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id BA2AB24963 for ; Mon, 4 May 2020 18:27:45 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=nvidia.com header.i=@nvidia.com header.b="hpPDYMTI" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1730821AbgEDS1o (ORCPT ); Mon, 4 May 2020 14:27:44 -0400 Received: from hqnvemgate24.nvidia.com ([216.228.121.143]:6452 "EHLO hqnvemgate24.nvidia.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1730700AbgEDS1o (ORCPT ); Mon, 4 May 2020 14:27:44 -0400 Received: from hqpgpgate101.nvidia.com (Not Verified[216.228.121.13]) by hqnvemgate24.nvidia.com (using TLS: TLSv1.2, DES-CBC3-SHA) id ; Mon, 04 May 2020 11:25:36 -0700 Received: from hqmail.nvidia.com ([172.20.161.6]) by hqpgpgate101.nvidia.com (PGP Universal service); Mon, 04 May 2020 11:27:43 -0700 X-PGP-Universal: processed; by hqpgpgate101.nvidia.com on Mon, 04 May 2020 11:27:43 -0700 Received: from HQMAIL105.nvidia.com (172.20.187.12) by HQMAIL101.nvidia.com (172.20.187.10) with Microsoft SMTP Server (TLS) id 15.0.1473.3; Mon, 4 May 2020 18:27:43 +0000 Received: from kwankhede-dev.nvidia.com (10.124.1.5) by HQMAIL105.nvidia.com (172.20.187.12) with Microsoft SMTP Server (TLS) id 15.0.1473.3 via Frontend Transport; Mon, 4 May 2020 18:27:36 +0000 From: Kirti Wankhede To: , CC: , , , , , , , , , , , , , , , , , , , , "Kirti Wankhede" Subject: [PATCH v1 1/2] Sample mtty: Add sysfs interface to pin pages Date: Mon, 4 May 2020 23:24:19 +0530 Message-ID: <1588614860-16330-2-git-send-email-kwankhede@nvidia.com> X-Mailer: git-send-email 2.7.0 In-Reply-To: <1588614860-16330-1-git-send-email-kwankhede@nvidia.com> References: <1588614860-16330-1-git-send-email-kwankhede@nvidia.com> X-NVConfidentiality: public MIME-Version: 1.0 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=nvidia.com; s=n1; t=1588616736; bh=LM7O/QtPrNroYPHMAw1tDGhJJm1iWHCvYUgOsUsGJ7U=; h=X-PGP-Universal:From:To:CC:Subject:Date:Message-ID:X-Mailer: In-Reply-To:References:X-NVConfidentiality:MIME-Version: Content-Type; b=hpPDYMTIWyA3RSFu+DgJWyk3YbCP/vjKSL74Gb5ZdospGbspIf2/V1Feypa9UgNrU aePDwyyivOy/Ul61M29DEr6mv+M0TWUs7YlRQKLdAZVcQ1ROs5FV/ciu60/Ht/P1a9 a9+p2Z6J2cG6zY0CSzOiqV73QCUFztxnN5KFzO2ymaDrnfIGGE7T4lzRcUcrH9dM7H yRAHOBcevjE9pUCPVU7GBlH87wfVnKxpP7+27dRC1J/xOrPVlkKA/IjIsS4ZzUap6y xreQplhyETU3I/mkFFnVKzWVkOkZ9M/cSc0NXReD93bmy8LkO9HDoi5V6mtoDp6QMi d9m1JVpb8j5QA== Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org Added sysfs interface to pin pages which will be used to test migration when vendor driver pins pages. Read on pin_pages shows number of pages pinned: $ cat /sys/bus/mdev/devices//vendor/pin_pages Pinned 0x0 pages Write gpfn to pin_pages to pin that page. One page pinned on write $ echo 0x20 > /sys/bus/mdev/devices//vendor/pin_pages Limitation: Buffer for 2GB system memory is created to track pinned pages. This is for testing purpose only. Signed-off-by: Kirti Wankhede --- samples/vfio-mdev/mtty.c | 176 ++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 175 insertions(+), 1 deletion(-) diff --git a/samples/vfio-mdev/mtty.c b/samples/vfio-mdev/mtty.c index ce84a300a4da..bf666cce5bb7 100644 --- a/samples/vfio-mdev/mtty.c +++ b/samples/vfio-mdev/mtty.c @@ -64,6 +64,10 @@ (((u64)(1) << MTTY_VFIO_PCI_OFFSET_SHIFT) - 1) #define MAX_MTTYS 24 +/* Maximum pages of 4K for upto 2G RAM can be pinned */ +#define MAX_GPFN_COUNT (512 * 1024) +#define PFN_NULL (~0UL) + /* * Global Structures */ @@ -141,6 +145,10 @@ struct mdev_state { struct mutex rxtx_lock; struct vfio_device_info dev_info; int nr_ports; + + /* List of pinned gpfns, gpfn as index and content is translated hpfn */ + unsigned long *gpfn_to_hpfn; + struct notifier_block nb; }; static struct mutex mdev_list_lock; @@ -745,6 +753,17 @@ static int mtty_create(struct kobject *kobj, struct mdev_device *mdev) return -ENOMEM; } + mdev_state->gpfn_to_hpfn = + kzalloc(sizeof(unsigned long) * MAX_GPFN_COUNT, GFP_KERNEL); + if (mdev_state->gpfn_to_hpfn == NULL) { + kfree(mdev_state->vconfig); + kfree(mdev_state); + return -ENOMEM; + } + + memset(mdev_state->gpfn_to_hpfn, ~0, + sizeof(unsigned long) * MAX_GPFN_COUNT); + mutex_init(&mdev_state->ops_lock); mdev_state->mdev = mdev; mdev_set_drvdata(mdev, mdev_state); @@ -769,6 +788,7 @@ static int mtty_remove(struct mdev_device *mdev) if (mdev_state == mds) { list_del(&mdev_state->next); mdev_set_drvdata(mdev, NULL); + kfree(mdev_state->gpfn_to_hpfn); kfree(mdev_state->vconfig); kfree(mdev_state); ret = 0; @@ -1246,15 +1266,95 @@ static long mtty_ioctl(struct mdev_device *mdev, unsigned int cmd, return -ENOTTY; } +static void unpin_pages_all(struct mdev_state *mdev_state) +{ + struct mdev_device *mdev = mdev_state->mdev; + unsigned long i; + + mutex_lock(&mdev_state->ops_lock); + for (i = 0; i < MAX_GPFN_COUNT; i++) { + if (mdev_state->gpfn_to_hpfn[i] != PFN_NULL) { + int ret; + + ret = vfio_unpin_pages(mdev_dev(mdev), &i, 1); + if (ret <= 0) { + pr_err("%s: 0x%lx unpin error %d\n", + __func__, i, ret); + continue; + } + mdev_state->gpfn_to_hpfn[i] = PFN_NULL; + } + } + mutex_unlock(&mdev_state->ops_lock); +} + +static int unmap_notifier(struct notifier_block *nb, unsigned long action, + void *data) +{ + if (action == VFIO_IOMMU_NOTIFY_DMA_UNMAP) { + struct mdev_state *mdev_state = container_of(nb, + struct mdev_state, nb); + struct mdev_device *mdev = mdev_state->mdev; + struct vfio_iommu_type1_dma_unmap *unmap = data; + unsigned long start = unmap->iova >> PAGE_SHIFT; + unsigned long end = (unmap->iova + unmap->size) >> PAGE_SHIFT; + unsigned long i; + + mutex_lock(&mdev_state->ops_lock); + for (i = start; i < end; i++) { + if (mdev_state->gpfn_to_hpfn[i] != PFN_NULL) { + int ret; + + ret = vfio_unpin_pages(mdev_dev(mdev), &i, 1); + if (ret <= 0) { + pr_err("%s: 0x%lx unpin error %d\n", + __func__, i, ret); + continue; + } + mdev_state->gpfn_to_hpfn[i] = PFN_NULL; + } + } + mutex_unlock(&mdev_state->ops_lock); + + } + return 0; +} + static int mtty_open(struct mdev_device *mdev) { + unsigned long events = VFIO_IOMMU_NOTIFY_DMA_UNMAP; + struct mdev_state *mdev_state; + int ret; + pr_info("%s\n", __func__); - return 0; + + if (!mdev) + return -EINVAL; + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return -ENODEV; + + mdev_state->nb.notifier_call = unmap_notifier; + + ret = vfio_register_notifier(mdev_dev(mdev), VFIO_IOMMU_NOTIFY, &events, + &mdev_state->nb); + return ret; } static void mtty_close(struct mdev_device *mdev) { + struct mdev_state *mdev_state; + pr_info("%s\n", __func__); + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return; + + unpin_pages_all(mdev_state); + vfio_unregister_notifier(mdev_dev(mdev), VFIO_IOMMU_NOTIFY, + &mdev_state->nb); } static ssize_t @@ -1293,8 +1393,82 @@ sample_mdev_dev_show(struct device *dev, struct device_attribute *attr, static DEVICE_ATTR_RO(sample_mdev_dev); +static ssize_t +pin_pages_show(struct device *dev, struct device_attribute *attr, char *buf) +{ + struct mdev_device *mdev = mdev_from_dev(dev); + struct mdev_state *mdev_state; + int i, count = 0; + + if (!mdev) + return -EINVAL; + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return -EINVAL; + + mutex_lock(&mdev_state->ops_lock); + for (i = 0; i < MAX_GPFN_COUNT; i++) { + if (mdev_state->gpfn_to_hpfn[i] != PFN_NULL) + count++; + } + mutex_unlock(&mdev_state->ops_lock); + return sprintf(buf, "Pinned 0x%x pages\n", count); +} + +static ssize_t +pin_pages_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + struct mdev_device *mdev = mdev_from_dev(dev); + struct mdev_state *mdev_state; + unsigned long gpfn, hpfn; + int ret; + + if (!mdev) + return -EINVAL; + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return -EINVAL; + + ret = kstrtoul(buf, 0, &gpfn); + if (ret) + return ret; + + if (gpfn >= MAX_GPFN_COUNT) { + pr_err("Error 0x%lx > 0x%lx\n", + gpfn, (unsigned long)MAX_GPFN_COUNT); + return -EINVAL; + } + + mutex_lock(&mdev_state->ops_lock); + + if (mdev_state->gpfn_to_hpfn[gpfn] != PFN_NULL) { + ret = -EEXIST; + goto out; + } + + ret = vfio_pin_pages(mdev_dev(mdev), &gpfn, 1, + IOMMU_READ | IOMMU_WRITE, &hpfn); + + if (ret <= 0) { + pr_err("Failed to pin, ret %d\n", ret); + goto out; + } + + mdev_state->gpfn_to_hpfn[gpfn] = hpfn; + ret = count; +out: + mutex_unlock(&mdev_state->ops_lock); + return ret; +} + +static DEVICE_ATTR_RW(pin_pages); + static struct attribute *mdev_dev_attrs[] = { &dev_attr_sample_mdev_dev.attr, + &dev_attr_pin_pages.attr, NULL, }; From patchwork Mon May 4 17:54:20 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Kirti Wankhede X-Patchwork-Id: 11527399 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id C0B4F17E8 for ; Mon, 4 May 2020 18:27:55 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id A382024953 for ; Mon, 4 May 2020 18:27:55 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=nvidia.com header.i=@nvidia.com header.b="psg4949v" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1730929AbgEDS1y (ORCPT ); Mon, 4 May 2020 14:27:54 -0400 Received: from hqnvemgate24.nvidia.com ([216.228.121.143]:6468 "EHLO hqnvemgate24.nvidia.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1730700AbgEDS1x (ORCPT ); Mon, 4 May 2020 14:27:53 -0400 Received: from hqpgpgate101.nvidia.com (Not Verified[216.228.121.13]) by hqnvemgate24.nvidia.com (using TLS: TLSv1.2, DES-CBC3-SHA) id ; Mon, 04 May 2020 11:25:44 -0700 Received: from hqmail.nvidia.com ([172.20.161.6]) by hqpgpgate101.nvidia.com (PGP Universal service); Mon, 04 May 2020 11:27:52 -0700 X-PGP-Universal: processed; by hqpgpgate101.nvidia.com on Mon, 04 May 2020 11:27:52 -0700 Received: from HQMAIL105.nvidia.com (172.20.187.12) by HQMAIL111.nvidia.com (172.20.187.18) with Microsoft SMTP Server (TLS) id 15.0.1473.3; Mon, 4 May 2020 18:27:51 +0000 Received: from kwankhede-dev.nvidia.com (10.124.1.5) by HQMAIL105.nvidia.com (172.20.187.12) with Microsoft SMTP Server (TLS) id 15.0.1473.3 via Frontend Transport; Mon, 4 May 2020 18:27:43 +0000 From: Kirti Wankhede To: , CC: , , , , , , , , , , , , , , , , , , , , "Kirti Wankhede" Subject: [PATCH v1 2/2] Sample mtty: Add migration capability to mtty module Date: Mon, 4 May 2020 23:24:20 +0530 Message-ID: <1588614860-16330-3-git-send-email-kwankhede@nvidia.com> X-Mailer: git-send-email 2.7.0 In-Reply-To: <1588614860-16330-1-git-send-email-kwankhede@nvidia.com> References: <1588614860-16330-1-git-send-email-kwankhede@nvidia.com> X-NVConfidentiality: public MIME-Version: 1.0 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=nvidia.com; s=n1; t=1588616744; bh=VQ98ApYetGzqpIkX7/EXIJ6vVyraG/3XX1ujqVvCTKE=; h=X-PGP-Universal:From:To:CC:Subject:Date:Message-ID:X-Mailer: In-Reply-To:References:X-NVConfidentiality:MIME-Version: Content-Type; b=psg4949v1Dz6LE+gM+r/drpyBqKieUGUxq6phXmRsJD94LY+U7e2mTLdniJ8L7i1K hXwV5H5ecszAVr5ZzBH59wMv6JVA9jw17Jz/DBKLRHWX1adNJ8bcnhjHaJv4Wlkp8v j6aGs2dIT/B60CmKT/jF9g1t33/iltWyJQeMsM+LBPME+ehgtnPM+0oUp9I8LJniHg 5UadFeye2OAWDxo7fmpfb20pEo9Pa2u4bEdlhqvvN/ZAyLP9j83fPCO+4Jvz4ZskWI 5Y5hz56YrkaF9TX7ZWBjQzuTAahOm4BboyMdn6I6sBxu84/JqOVJGOySxPOdCl9vUY of8oukkosC0SQ== Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org This patch makes mtty device migration capable. Purpose od this code is to test migration interface. Only stop-and-copy phase is implemented. Postcopy migration is not supported. Actual data for mtty device migration is very less. Appended dummy data to migration data stream, default 100 Mbytes. Added sysfs file 'dummy_data_size_MB' to get dummy data size from user which can be used to check performance of based of data size. During resuming dummy data is read and discarded. Signed-off-by: Kirti Wankhede --- samples/vfio-mdev/mtty.c | 602 ++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 574 insertions(+), 28 deletions(-) diff --git a/samples/vfio-mdev/mtty.c b/samples/vfio-mdev/mtty.c index bf666cce5bb7..f9194234fc6a 100644 --- a/samples/vfio-mdev/mtty.c +++ b/samples/vfio-mdev/mtty.c @@ -44,9 +44,23 @@ #define MTTY_STRING_LEN 16 -#define MTTY_CONFIG_SPACE_SIZE 0xff -#define MTTY_IO_BAR_SIZE 0x8 -#define MTTY_MMIO_BAR_SIZE 0x100000 +#define MTTY_CONFIG_SPACE_SIZE 0xff +#define MTTY_IO_BAR_SIZE 0x8 +#define MTTY_MMIO_BAR_SIZE 0x100000 +#define MTTY_MIGRATION_REGION_SIZE 0x1000000 // 16M + +#define MTTY_MIGRATION_REGION_INDEX VFIO_PCI_NUM_REGIONS +#define MTTY_REGIONS_MAX (MTTY_MIGRATION_REGION_INDEX + 1) + +/* Data section start from page aligned offset */ +#define MTTY_MIGRATION_REGION_DATA_OFFSET (0x1000) + +/* First page is used for struct vfio_device_migration_info */ +#define MTTY_MIGRATION_REGION_SIZE_MMAP \ + (MTTY_MIGRATION_REGION_SIZE - MTTY_MIGRATION_REGION_DATA_OFFSET) + +#define MIGRATION_INFO_OFFSET(MEMBER) \ + offsetof(struct vfio_device_migration_info, MEMBER) #define STORE_LE16(addr, val) (*(u16 *)addr = val) #define STORE_LE32(addr, val) (*(u32 *)addr = val) @@ -129,6 +143,28 @@ struct serial_port { u8 intr_trigger_level; /* interrupt trigger level */ }; +/* Migration packet */ +#define PACKET_ID (u16)(0xfeedbaba) + +#define PACKET_FLAGS_ACTUAL_DATA (1 << 0) +#define PACKET_FLAGS_DUMMY_DATA (1 << 1) + +#define PACKET_DATA_SIZE_MAX (8 * 1024 * 1024) + +struct packet { + u16 id; + u16 flags; + u32 data_size; + u8 data[]; +}; + +enum { + PACKET_STATE_NONE = 0, + PACKET_STATE_PREPARED, + PACKET_STATE_COPIED, + PACKET_STATE_LAST, +}; + /* State of each mdev device */ struct mdev_state { int irq_fd; @@ -138,22 +174,37 @@ struct mdev_state { u8 *vconfig; struct mutex ops_lock; struct mdev_device *mdev; - struct mdev_region_info region_info[VFIO_PCI_NUM_REGIONS]; - u32 bar_mask[VFIO_PCI_NUM_REGIONS]; + struct mdev_region_info region_info[MTTY_REGIONS_MAX]; + u32 bar_mask[MTTY_REGIONS_MAX]; struct list_head next; struct serial_port s[2]; struct mutex rxtx_lock; struct vfio_device_info dev_info; - int nr_ports; + u32 nr_ports; /* List of pinned gpfns, gpfn as index and content is translated hpfn */ unsigned long *gpfn_to_hpfn; struct notifier_block nb; + + u32 device_state; + u64 saved_size; + void *mig_region_base; + bool is_actual_data_sent; + struct packet *pkt; + u32 packet_state; + u64 dummy_data_size; }; static struct mutex mdev_list_lock; static struct list_head mdev_devices_list; +/* + * Default dummy data size set to 100 MB. To change value of dummy data size at + * runtime but before migration write size in MB to sysfs file + * dummy_data_size_MB + */ +static unsigned long user_dummy_data_size = (100 * 1024 * 1024); + static const struct file_operations vd_fops = { .owner = THIS_MODULE, }; @@ -639,6 +690,288 @@ static void mdev_read_base(struct mdev_state *mdev_state) } } +static int save_setup(struct mdev_state *mdev_state) +{ + mdev_state->is_actual_data_sent = false; + + memset(mdev_state->pkt, 0, sizeof(struct packet) + + PACKET_DATA_SIZE_MAX); + + return 0; +} + +static int set_device_state(struct mdev_state *mdev_state, u32 device_state) +{ + int ret = 0; + + if (mdev_state->device_state == device_state) + return 0; + + if (device_state & VFIO_DEVICE_STATE_RUNNING) { +#if defined(DEBUG) + if (device_state & VFIO_DEVICE_STATE_SAVING) { + pr_info("%s: %s Pre-copy\n", __func__, + dev_name(mdev_dev(mdev_state->mdev))); + } else + pr_info("%s: %s Running\n", __func__, + dev_name(mdev_dev(mdev_state->mdev))); +#endif + } else { + if (device_state & VFIO_DEVICE_STATE_SAVING) { +#if defined(DEBUG) + pr_info("%s: %s Stop-n-copy\n", __func__, + dev_name(mdev_dev(mdev_state->mdev))); +#endif + ret = save_setup(mdev_state); + + } else if (device_state & VFIO_DEVICE_STATE_RESUMING) { +#if defined(DEBUG) + pr_info("%s: %s Resuming\n", __func__, + dev_name(mdev_dev(mdev_state->mdev))); + } else { + pr_info("%s: %s Stopped\n", __func__, + dev_name(mdev_dev(mdev_state->mdev))); +#endif + } + } + + mdev_state->device_state = device_state; + + return ret; +} + +static u32 get_device_state(struct mdev_state *mdev_state) +{ + return mdev_state->device_state; +} + +static void write_to_packet(struct packet *pkt, u8 *data, size_t size) +{ + if ((pkt->data_size + size) > PACKET_DATA_SIZE_MAX) { + pr_err("%s: packet data overflow\n", __func__); + return; + } + memcpy((void *)&pkt->data[pkt->data_size], (void *)data, size); + pkt->data_size += size; +} + +static void read_from_packet(struct packet *pkt, u8 *data, + int index, size_t size) +{ + if ((index + size) > PACKET_DATA_SIZE_MAX) { + pr_err("%s: packet data overflow\n", __func__); + return; + } + + memcpy((void *)data, (void *)&pkt->data[index], size); +} + +static int save_device_data(struct mdev_state *mdev_state, u64 *pending) +{ + /* Save device data only during stop-and-copy phase */ + if (mdev_state->device_state != VFIO_DEVICE_STATE_SAVING) { + *pending = 0; + return 0; + } + + if (mdev_state->packet_state == PACKET_STATE_PREPARED) { + *pending = sizeof(struct packet) + mdev_state->pkt->data_size; + return 0; + } + + if (!mdev_state->is_actual_data_sent) { + + /* create actual data packet */ + write_to_packet(mdev_state->pkt, (u8 *)&mdev_state->nr_ports, + sizeof(mdev_state->nr_ports)); + write_to_packet(mdev_state->pkt, (u8 *)&mdev_state->s, + sizeof(struct serial_port) * 2); + + write_to_packet(mdev_state->pkt, mdev_state->vconfig, + MTTY_CONFIG_SPACE_SIZE); + + write_to_packet(mdev_state->pkt, (u8 *)mdev_state->gpfn_to_hpfn, + sizeof(unsigned long) * MAX_GPFN_COUNT); + + mdev_state->pkt->id = PACKET_ID; + mdev_state->pkt->flags = PACKET_FLAGS_ACTUAL_DATA; + + mdev_state->is_actual_data_sent = true; + } else { + /* create dummy data packet */ + if (mdev_state->dummy_data_size > user_dummy_data_size) { + *pending = 0; + mdev_state->packet_state = PACKET_STATE_NONE; + return 0; + } + + memset(mdev_state->pkt->data, 0xa5, PACKET_DATA_SIZE_MAX); + + mdev_state->pkt->id = PACKET_ID; + mdev_state->pkt->flags = PACKET_FLAGS_DUMMY_DATA; + mdev_state->pkt->data_size = PACKET_DATA_SIZE_MAX; + mdev_state->dummy_data_size += PACKET_DATA_SIZE_MAX; + } + + *pending = sizeof(struct packet) + mdev_state->pkt->data_size; + mdev_state->packet_state = PACKET_STATE_PREPARED; + mdev_state->saved_size = 0; + + return 0; +} + +static int copy_device_data(struct mdev_state *mdev_state) +{ + u64 size; + + if (!mdev_state->pkt || !mdev_state->mig_region_base) + return -EINVAL; + + if (mdev_state->packet_state == PACKET_STATE_COPIED) + return 0; + + if (!mdev_state->pkt->data_size) + return 0; + + size = sizeof(struct packet) + mdev_state->pkt->data_size; + + memcpy(mdev_state->mig_region_base, mdev_state->pkt, size); + + mdev_state->saved_size = size; + mdev_state->packet_state = PACKET_STATE_COPIED; + memset(mdev_state->pkt, 0, sizeof(struct packet)); + return 0; +} + +static int resume_device_data(struct mdev_state *mdev_state, u64 data_size) +{ + unsigned long i; + + if (mdev_state->device_state != VFIO_DEVICE_STATE_RESUMING) + return -EINVAL; + + if (!mdev_state->pkt || !mdev_state->mig_region_base) + return -EINVAL; + + memcpy(mdev_state->pkt, mdev_state->mig_region_base, data_size); + + if (mdev_state->pkt->flags & PACKET_FLAGS_ACTUAL_DATA) { + int index = 0; + /* restore device data */ + read_from_packet(mdev_state->pkt, (u8 *)&mdev_state->nr_ports, + index, sizeof(mdev_state->nr_ports)); + index += sizeof(mdev_state->nr_ports); + + read_from_packet(mdev_state->pkt, (u8 *)&mdev_state->s, + index, sizeof(struct serial_port) * 2); + index += sizeof(struct serial_port) * 2; + + read_from_packet(mdev_state->pkt, mdev_state->vconfig, + index, MTTY_CONFIG_SPACE_SIZE); + index += MTTY_CONFIG_SPACE_SIZE; + + read_from_packet(mdev_state->pkt, + (u8 *)mdev_state->gpfn_to_hpfn, + index, sizeof(unsigned long) * MAX_GPFN_COUNT); + index += sizeof(unsigned long) * MAX_GPFN_COUNT; + + for (i = 0; i < MAX_GPFN_COUNT; i++) { + if (mdev_state->gpfn_to_hpfn[i] != PFN_NULL) { + int ret; + unsigned long hpfn; + + ret = vfio_pin_pages(mdev_dev(mdev_state->mdev), + &i, 1, IOMMU_READ | IOMMU_WRITE, &hpfn); + if (ret <= 0) { + pr_err("%s: 0x%lx unpin error %d\n", + __func__, i, ret); + continue; + } + mdev_state->gpfn_to_hpfn[i] = hpfn; + } + } + } else { +#if defined(DEBUG) + pr_info("%s: %s discard data 0x%llx\n", + __func__, dev_name(mdev_dev(mdev_state->mdev)), + data_size); +#endif + } + + return 0; +} + +static int handle_mig_read(unsigned int index, struct mdev_state *mdev_state, + loff_t offset, u8 *buf, u32 count) +{ + int ret = 0; + u64 pending = 0; + + switch (offset) { + case MIGRATION_INFO_OFFSET(device_state): // 0x00 + *(u32 *)buf = get_device_state(mdev_state); + break; + + case MIGRATION_INFO_OFFSET(pending_bytes): // 0x08 + ret = save_device_data(mdev_state, &pending); + if (ret) + break; + *(u64 *)buf = pending; + break; + + case MIGRATION_INFO_OFFSET(data_offset): // 0x10 + if (mdev_state->device_state & VFIO_DEVICE_STATE_SAVING) { + ret = copy_device_data(mdev_state); + if (ret) + break; + } + *(u64 *)buf = MTTY_MIGRATION_REGION_DATA_OFFSET; + break; + + case MIGRATION_INFO_OFFSET(data_size): // 0x18 + *(u64 *)buf = mdev_state->saved_size; + break; + + default: + ret = -EINVAL; + } + +#if defined(DEBUG) + pr_info("%s: %s MIG RD @0x%llx bytes: %d data: 0x%x\n", + __func__, dev_name(mdev_dev(mdev_state->mdev)), + offset, count, *(u32 *)buf); +#endif + return ret; +} + +static int handle_mig_write(unsigned int index, struct mdev_state *mdev_state, + loff_t offset, u8 *buf, u32 count) +{ + int ret = 0; + +#if defined(DEBUG) + pr_info("%s: %s MIG WR @0x%llx bytes: %d data: 0x%x\n", + __func__, dev_name(mdev_dev(mdev_state->mdev)), + offset, count, *(u32 *)buf); +#endif + switch (offset) { + case MIGRATION_INFO_OFFSET(device_state): // 0x00 + ret = set_device_state(mdev_state, *(u32 *)buf); + break; + + case MIGRATION_INFO_OFFSET(data_size): // 0x18 + ret = resume_device_data(mdev_state, *(u64 *)buf); + break; + + case MIGRATION_INFO_OFFSET(pending_bytes): // 0x08 + case MIGRATION_INFO_OFFSET(data_offset): // 0x10 + default: + ret = -EINVAL; + } + + return ret; +} + static ssize_t mdev_access(struct mdev_device *mdev, u8 *buf, size_t count, loff_t pos, bool is_write) { @@ -702,6 +1035,18 @@ static ssize_t mdev_access(struct mdev_device *mdev, u8 *buf, size_t count, } break; + case MTTY_MIGRATION_REGION_INDEX: + if (is_write) { + ret = handle_mig_write(index, mdev_state, offset, buf, + count); + } else { + ret = handle_mig_read(index, mdev_state, offset, buf, + count); + } + if (ret) + goto accessfailed; + break; + default: ret = -1; goto accessfailed; @@ -709,7 +1054,6 @@ static ssize_t mdev_access(struct mdev_device *mdev, u8 *buf, size_t count, ret = count; - accessfailed: mutex_unlock(&mdev_state->ops_lock); @@ -819,13 +1163,29 @@ static int mtty_reset(struct mdev_device *mdev) static ssize_t mtty_read(struct mdev_device *mdev, char __user *buf, size_t count, loff_t *ppos) { - unsigned int done = 0; + unsigned int done = 0, index; int ret; + index = MTTY_VFIO_PCI_OFFSET_TO_INDEX(*ppos); + while (count) { size_t filled; - if (count >= 4 && !(*ppos % 4)) { + if ((index == MTTY_MIGRATION_REGION_INDEX) && + (count >= 8 && !(*ppos % 8))) { + u64 val; + + ret = mdev_access(mdev, (u8 *)&val, sizeof(val), + *ppos, false); + if (ret <= 0) + goto read_err; + + if (copy_to_user(buf, &val, sizeof(val))) + goto read_err; + + filled = 8; + + } else if (count >= 4 && !(*ppos % 4)) { u32 val; ret = mdev_access(mdev, (u8 *)&val, sizeof(val), @@ -878,13 +1238,27 @@ static ssize_t mtty_read(struct mdev_device *mdev, char __user *buf, static ssize_t mtty_write(struct mdev_device *mdev, const char __user *buf, size_t count, loff_t *ppos) { - unsigned int done = 0; + unsigned int done = 0, index; int ret; + index = MTTY_VFIO_PCI_OFFSET_TO_INDEX(*ppos); while (count) { size_t filled; - if (count >= 4 && !(*ppos % 4)) { + if ((index == MTTY_MIGRATION_REGION_INDEX) && + (count >= 8 && !(*ppos % 8))) { + u64 val; + + if (copy_from_user(&val, buf, sizeof(val))) + goto write_err; + + ret = mdev_access(mdev, (u8 *)&val, sizeof(val), + *ppos, true); + if (ret <= 0) + goto write_err; + + filled = 8; + } else if (count >= 4 && !(*ppos % 4)) { u32 val; if (copy_from_user(&val, buf, sizeof(val))) @@ -1061,12 +1435,13 @@ static int mtty_trigger_interrupt(struct mdev_state *mdev_state) } static int mtty_get_region_info(struct mdev_device *mdev, - struct vfio_region_info *region_info, - u16 *cap_type_id, void **cap_type) + struct vfio_region_info *region_info, + struct vfio_info_cap *caps) { unsigned int size = 0; struct mdev_state *mdev_state; - u32 bar_index; + u32 index; + int ret = 0; if (!mdev) return -EINVAL; @@ -1075,13 +1450,13 @@ static int mtty_get_region_info(struct mdev_device *mdev, if (!mdev_state) return -EINVAL; - bar_index = region_info->index; - if (bar_index >= VFIO_PCI_NUM_REGIONS) + index = region_info->index; + if (index >= MTTY_REGIONS_MAX) return -EINVAL; mutex_lock(&mdev_state->ops_lock); - switch (bar_index) { + switch (index) { case VFIO_PCI_CONFIG_REGION_INDEX: size = MTTY_CONFIG_SPACE_SIZE; break; @@ -1092,21 +1467,63 @@ static int mtty_get_region_info(struct mdev_device *mdev, if (mdev_state->nr_ports == 2) size = MTTY_IO_BAR_SIZE; break; + case MTTY_MIGRATION_REGION_INDEX: + size = MTTY_MIGRATION_REGION_SIZE; + break; default: size = 0; break; } - mdev_state->region_info[bar_index].size = size; - mdev_state->region_info[bar_index].vfio_offset = - MTTY_VFIO_PCI_INDEX_TO_OFFSET(bar_index); + mdev_state->region_info[index].size = size; + mdev_state->region_info[index].vfio_offset = + MTTY_VFIO_PCI_INDEX_TO_OFFSET(index); region_info->size = size; - region_info->offset = MTTY_VFIO_PCI_INDEX_TO_OFFSET(bar_index); + region_info->offset = MTTY_VFIO_PCI_INDEX_TO_OFFSET(index); region_info->flags = VFIO_REGION_INFO_FLAG_READ | - VFIO_REGION_INFO_FLAG_WRITE; + VFIO_REGION_INFO_FLAG_WRITE; + + if (index == MTTY_MIGRATION_REGION_INDEX) { + struct vfio_region_info_cap_sparse { + struct vfio_region_info_cap_sparse_mmap sparse; + struct vfio_region_sparse_mmap_area area; + }; + + struct vfio_region_info_cap_sparse mig_region; + + struct vfio_region_info_cap_type cap_type = { + .header.id = VFIO_REGION_INFO_CAP_TYPE, + .header.version = 1, + .type = VFIO_REGION_TYPE_MIGRATION, + .subtype = VFIO_REGION_SUBTYPE_MIGRATION + }; + + /* Add REGION CAP type */ + ret = vfio_info_add_capability(caps, &cap_type.header, + sizeof(cap_type)); + if (ret) + goto exit; + + /* Add sparse mmap cap type */ + mig_region.sparse.nr_areas = 1; + mig_region.sparse.header.id = VFIO_REGION_INFO_CAP_SPARSE_MMAP; + mig_region.sparse.header.version = 1; + + mig_region.area.offset = MTTY_MIGRATION_REGION_DATA_OFFSET; + mig_region.area.size = MTTY_MIGRATION_REGION_SIZE_MMAP; + + region_info->flags |= VFIO_REGION_INFO_FLAG_CAPS; + + if (region_info->argsz > sizeof(*region_info)) + region_info->flags |= VFIO_REGION_INFO_FLAG_MMAP; + + ret = vfio_info_add_capability(caps, &mig_region.sparse.header, + sizeof(mig_region)); + } +exit: mutex_unlock(&mdev_state->ops_lock); - return 0; + return ret; } static int mtty_get_irq_info(struct mdev_device *mdev, @@ -1138,7 +1555,7 @@ static int mtty_get_device_info(struct mdev_device *mdev, struct vfio_device_info *dev_info) { dev_info->flags = VFIO_DEVICE_FLAGS_PCI; - dev_info->num_regions = VFIO_PCI_NUM_REGIONS; + dev_info->num_regions = MTTY_REGIONS_MAX; dev_info->num_irqs = VFIO_PCI_NUM_IRQS; return 0; @@ -1150,6 +1567,7 @@ static long mtty_ioctl(struct mdev_device *mdev, unsigned int cmd, int ret = 0; unsigned long minsz; struct mdev_state *mdev_state; + struct vfio_info_cap caps = { .buf = NULL, .size = 0 }; if (!mdev) return -EINVAL; @@ -1185,8 +1603,6 @@ static long mtty_ioctl(struct mdev_device *mdev, unsigned int cmd, case VFIO_DEVICE_GET_REGION_INFO: { struct vfio_region_info info; - u16 cap_type_id = 0; - void *cap_type = NULL; minsz = offsetofend(struct vfio_region_info, offset); @@ -1196,11 +1612,29 @@ static long mtty_ioctl(struct mdev_device *mdev, unsigned int cmd, if (info.argsz < minsz) return -EINVAL; - ret = mtty_get_region_info(mdev, &info, &cap_type_id, - &cap_type); + ret = mtty_get_region_info(mdev, &info, &caps); if (ret) return ret; + if (caps.size) { + info.flags |= VFIO_REGION_INFO_FLAG_CAPS; + if (info.argsz < sizeof(info) + caps.size) { + info.argsz = sizeof(info) + caps.size; + info.cap_offset = 0; + } else { + vfio_info_cap_shift(&caps, sizeof(info)); + if (copy_to_user((void __user *)arg + + sizeof(info), caps.buf, + caps.size)) { + kfree(caps.buf); + ret = -EFAULT; + break; + } + info.cap_offset = sizeof(info); + } + kfree(caps.buf); + } + if (copy_to_user((void __user *)arg, &info, minsz)) return -EFAULT; @@ -1266,6 +1700,89 @@ static long mtty_ioctl(struct mdev_device *mdev, unsigned int cmd, return -ENOTTY; } +void mmap_close(struct vm_area_struct *vma) +{ + struct mdev_device *mdev = vma->vm_private_data; + struct mdev_state *mdev_state; + uint32_t index = 0; + + if (!mdev) + return; + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return; + + mutex_lock(&mdev_state->ops_lock); + index = MTTY_VFIO_PCI_OFFSET_TO_INDEX(vma->vm_pgoff << PAGE_SHIFT); + if (index == MTTY_MIGRATION_REGION_INDEX) { + if (mdev_state->mig_region_base != NULL) { + vfree(mdev_state->mig_region_base); + mdev_state->mig_region_base = NULL; + } + + if (mdev_state->pkt != NULL) { + vfree(mdev_state->pkt); + mdev_state->pkt = NULL; + } + } + mutex_unlock(&mdev_state->ops_lock); +} + +static const struct vm_operations_struct mdev_vm_ops = { + .close = mmap_close, +}; + +static int mtty_mmap(struct mdev_device *mdev, struct vm_area_struct *vma) +{ + struct mdev_state *mdev_state; + unsigned int index; + int ret = 0; + + if (!mdev) + return -EINVAL; + + mdev_state = mdev_get_drvdata(mdev); + if (!mdev_state) + return -ENODEV; + + mutex_lock(&mdev_state->ops_lock); + + index = MTTY_VFIO_PCI_OFFSET_TO_INDEX(vma->vm_pgoff << PAGE_SHIFT); + if (index == MTTY_MIGRATION_REGION_INDEX) { + mdev_state->mig_region_base = + vmalloc_user(MTTY_MIGRATION_REGION_SIZE_MMAP); + if (mdev_state->mig_region_base == NULL) { + ret = -ENOMEM; + goto mmap_exit; + } + + mdev_state->pkt = vzalloc(sizeof(struct packet) + + PACKET_DATA_SIZE_MAX); + if (mdev_state->pkt == NULL) { + vfree(mdev_state->mig_region_base); + mdev_state->mig_region_base = NULL; + ret = -ENOMEM; + goto mmap_exit; + } + + vma->vm_ops = &mdev_vm_ops; + + ret = remap_vmalloc_range(vma, mdev_state->mig_region_base, 0); + if (ret != 0) { + pr_err("remap_vmalloc_range failed, ret= %d\n", ret); + vfree(mdev_state->mig_region_base); + mdev_state->mig_region_base = NULL; + vfree(mdev_state->pkt); + mdev_state->pkt = NULL; + goto mmap_exit; + } + } +mmap_exit: + mutex_unlock(&mdev_state->ops_lock); + return ret; +} + static void unpin_pages_all(struct mdev_state *mdev_state) { struct mdev_device *mdev = mdev_state->mdev; @@ -1339,6 +1856,8 @@ static int mtty_open(struct mdev_device *mdev) ret = vfio_register_notifier(mdev_dev(mdev), VFIO_IOMMU_NOTIFY, &events, &mdev_state->nb); + mdev_state->dummy_data_size = 0; + mdev_state->mig_region_base = NULL; return ret; } @@ -1355,6 +1874,15 @@ static void mtty_close(struct mdev_device *mdev) unpin_pages_all(mdev_state); vfio_unregister_notifier(mdev_dev(mdev), VFIO_IOMMU_NOTIFY, &mdev_state->nb); + if (mdev_state->pkt != NULL) { + vfree(mdev_state->pkt); + mdev_state->pkt = NULL; + } + + if (mdev_state->mig_region_base != NULL) { + vfree(mdev_state->mig_region_base); + mdev_state->mig_region_base = NULL; + } } static ssize_t @@ -1466,9 +1994,26 @@ pin_pages_store(struct device *dev, struct device_attribute *attr, static DEVICE_ATTR_RW(pin_pages); +static ssize_t +dummy_data_size_MB_store(struct device *dev, struct device_attribute *attr, + const char *buf, size_t count) +{ + int ret; + + ret = kstrtoul(buf, 0, &user_dummy_data_size); + if (ret) + return ret; + + user_dummy_data_size = user_dummy_data_size << 20; + return count; +} + +static DEVICE_ATTR_WO(dummy_data_size_MB); + static struct attribute *mdev_dev_attrs[] = { &dev_attr_sample_mdev_dev.attr, &dev_attr_pin_pages.attr, + &dev_attr_dummy_data_size_MB.attr, NULL, }; @@ -1573,6 +2118,7 @@ static const struct mdev_parent_ops mdev_fops = { .read = mtty_read, .write = mtty_write, .ioctl = mtty_ioctl, + .mmap = mtty_mmap, }; static void mtty_device_release(struct device *dev)