DPDK patches and discussions
 help / color / mirror / Atom feed
From: "谢华伟(此时此刻)" <huawei.xhw@alibaba-inc.com>
To: dev@dpdk.org
Cc: ferruh.yigit@intel.com, zhihong.wang@intel.com, chenbo.xia@intel.com
Subject: [dpdk-dev] [PATCH] pci: support both PIO and MMIO bar for legacy virtio on x86
Date: Mon, 14 Sep 2020 20:43:12 +0800	[thread overview]
Message-ID: <598a88dc-37c7-b525-9ab7-21f33415bfa2@alibaba-inc.com> (raw)

In previous implementation, with igb_uio we get PIO address from igb_uio
sysfs entry; with uio_pci_generic, we get PIO address from
/proc/ioports.
For PIO/MMIO RW, there is different path for different drivers and arch.
For VFIO, PIO port RW is through syscall, which has big performance issue.
On X86, it assumes only PIO is supported.

This is too much twisted.
This patch unify the way to get both PIO and MMIO address for different 
drivers and arch, all from standard resource attr under pci sysfs.

Only configure space rw should go through driver specific method.

Signed-off-by: huawei.xhw <huawei.xhw@alibaba-inc.com>
---
  drivers/bus/pci/linux/pci.c     |  77 +-------------------
  drivers/bus/pci/linux/pci_uio.c | 156 
+++++++++++++++++++++++++++-------------
  2 files changed, 110 insertions(+), 123 deletions(-)

diff --git a/drivers/bus/pci/linux/pci.c b/drivers/bus/pci/linux/pci.c
index a2198ab..7266001 100644
--- a/drivers/bus/pci/linux/pci.c
+++ b/drivers/bus/pci/linux/pci.c
@@ -687,71 +687,6 @@ int rte_pci_write_config(const struct 
rte_pci_device *device,
      }
  }

-#if defined(RTE_ARCH_X86)
-static int
-pci_ioport_map(struct rte_pci_device *dev, int bar __rte_unused,
-        struct rte_pci_ioport *p)
-{
-    uint16_t start, end;
-    FILE *fp;
-    char *line = NULL;
-    char pci_id[16];
-    int found = 0;
-    size_t linesz;
-
-    if (rte_eal_iopl_init() != 0) {
-        RTE_LOG(ERR, EAL, "%s(): insufficient ioport permissions for 
PCI device %s\n",
-            __func__, dev->name);
-        return -1;
-    }
-
-    snprintf(pci_id, sizeof(pci_id), PCI_PRI_FMT,
-         dev->addr.domain, dev->addr.bus,
-         dev->addr.devid, dev->addr.function);
-
-    fp = fopen("/proc/ioports", "r");
-    if (fp == NULL) {
-        RTE_LOG(ERR, EAL, "%s(): can't open ioports\n", __func__);
-        return -1;
-    }
-
-    while (getdelim(&line, &linesz, '\n', fp) > 0) {
-        char *ptr = line;
-        char *left;
-        int n;
-
-        n = strcspn(ptr, ":");
-        ptr[n] = 0;
-        left = &ptr[n + 1];
-
-        while (*left && isspace(*left))
-            left++;
-
-        if (!strncmp(left, pci_id, strlen(pci_id))) {
-            found = 1;
-
-            while (*ptr && isspace(*ptr))
-                ptr++;
-
-            sscanf(ptr, "%04hx-%04hx", &start, &end);
-
-            break;
-        }
-    }
-
-    free(line);
-    fclose(fp);
-
-    if (!found)
-        return -1;
-
-    p->base = start;
-    RTE_LOG(DEBUG, EAL, "PCI Port IO found start=0x%x\n", start);
-
-    return 0;
-}
-#endif
-
  int
  rte_pci_ioport_map(struct rte_pci_device *dev, int bar,
          struct rte_pci_ioport *p)
@@ -762,18 +697,12 @@ int rte_pci_write_config(const struct 
rte_pci_device *device,
  #ifdef VFIO_PRESENT
      case RTE_KDRV_VFIO:
          if (pci_vfio_is_enabled())
-            ret = pci_vfio_ioport_map(dev, bar, p);
+            ret = pci_uio_ioport_map(dev, bar, p);
          break;
  #endif
      case RTE_KDRV_IGB_UIO:
-        ret = pci_uio_ioport_map(dev, bar, p);
-        break;
      case RTE_KDRV_UIO_GENERIC:
-#if defined(RTE_ARCH_X86)
-        ret = pci_ioport_map(dev, bar, p);
-#else
          ret = pci_uio_ioport_map(dev, bar, p);
-#endif
          break;
      default:
          break;
@@ -792,7 +721,7 @@ int rte_pci_write_config(const struct rte_pci_device 
*device,
      switch (p->dev->kdrv) {
  #ifdef VFIO_PRESENT
      case RTE_KDRV_VFIO:
-        pci_vfio_ioport_read(p, data, len, offset);
+        pci_uio_ioport_read(p, data, len, offset);
          break;
  #endif
      case RTE_KDRV_IGB_UIO:
@@ -813,7 +742,7 @@ int rte_pci_write_config(const struct rte_pci_device 
*device,
      switch (p->dev->kdrv) {
  #ifdef VFIO_PRESENT
      case RTE_KDRV_VFIO:
-        pci_vfio_ioport_write(p, data, len, offset);
+        pci_uio_ioport_write(p, data, len, offset);
          break;
  #endif
      case RTE_KDRV_IGB_UIO:
diff --git a/drivers/bus/pci/linux/pci_uio.c 
b/drivers/bus/pci/linux/pci_uio.c
index 097dc19..0d7ee3c 100644
--- a/drivers/bus/pci/linux/pci_uio.c
+++ b/drivers/bus/pci/linux/pci_uio.c
@@ -372,36 +372,60 @@
  pci_uio_ioport_map(struct rte_pci_device *dev, int bar,
             struct rte_pci_ioport *p)
  {
+    FILE *f;
      char dirname[PATH_MAX];
      char filename[PATH_MAX];
+    char buf[BUFSIZ];
+    uint64_t phys_addr, end_addr, flags;
      int uio_num;
-    unsigned long start;
+    unsigned long base;
+    bool iobar;
+    int i;

-    if (rte_eal_iopl_init() != 0) {
-        RTE_LOG(ERR, EAL, "%s(): insufficient ioport permissions for 
PCI device %s\n",
-            __func__, dev->name);
+    /* open and read addresses of the corresponding resource in sysfs */
+    snprintf(filename, sizeof(filename), "%s/" PCI_PRI_FMT "/resource",
+        rte_pci_get_sysfs_path(), dev->addr.domain, dev->addr.bus,
+        dev->addr.devid, dev->addr.function);
+    f = fopen(filename, "r");
+    if (f == NULL) {
+        RTE_LOG(ERR, EAL, "Cannot open sysfs resource: %s\n",
+            strerror(errno));
          return -1;
      }
+    for (i = 0; i < bar + 1; i++) {
+        if (fgets(buf, sizeof(buf), f) == NULL) {
+            RTE_LOG(ERR, EAL, "Cannot read sysfs resource\n");
+            goto error;
+        }
+    }
+    if (pci_parse_one_sysfs_resource(buf, sizeof(buf), &phys_addr,
+            &end_addr, &flags) < 0)
+        goto error;

-    uio_num = pci_get_uio_dev(dev, dirname, sizeof(dirname), 0);
-    if (uio_num < 0)
-        return -1;
+    if (flags & IORESOURCE_IO) {
+        iobar = 1;
+        base = (unsigned long)phys_addr;
+        RTE_LOG(INFO, EAL, "%s(): iobar %08lx detected\n", __func__, base);
+    } else if (flags & IORESOURCE_MEM) {
+        iobar = 0;
+        base = (unsigned long)dev->mem_resource[bar].addr;
+        RTE_LOG(INFO, EAL, "%s(): membar %08lx detected\n", __func__, 
base);
+    } else {
+        RTE_LOG(ERR, EAL, "%s(): unknown bar type\n", __func__);
+        goto error;
+    }

-    /* get portio start */
-    snprintf(filename, sizeof(filename),
-         "%s/portio/port%d/start", dirname, bar);
-    if (eal_parse_sysfs_value(filename, &start) < 0) {
-        RTE_LOG(ERR, EAL, "%s(): cannot parse portio start\n",
-            __func__);
+    if (iobar && rte_eal_iopl_init() != 0) {
+        RTE_LOG(ERR, EAL, "%s(): insufficient ioport permissions for 
PCI device %s\n",
+            __func__, dev->name);
          return -1;
      }
-    /* ensure we don't get anything funny here, read/write will cast to
-     * uin16_t */
-    if (start > UINT16_MAX)
-        return -1;

      /* FIXME only for primary process ? */
      if (dev->intr_handle.type == RTE_INTR_HANDLE_UNKNOWN) {
+        uio_num = pci_get_uio_dev(dev, dirname, sizeof(dirname), 0);
+        if (uio_num < 0)
+            return -1;

          snprintf(filename, sizeof(filename), "/dev/uio%u", uio_num);
          dev->intr_handle.fd = open(filename, O_RDWR);
@@ -413,11 +437,14 @@
          dev->intr_handle.type = RTE_INTR_HANDLE_UIO;
      }

-    RTE_LOG(DEBUG, EAL, "PCI Port IO found start=0x%lx\n", start);
+    RTE_LOG(DEBUG, EAL, "PCI Port IO found start=0x%lx\n", base);

-    p->base = start;
+    p->base = base;
      p->len = 0;
      return 0;
+error:
+    fclose(f);
+    return -1;
  }
  #else
  int
@@ -488,6 +515,61 @@
  }
  #endif

+#define PIO_MAX 0x10000
+static inline uint8_t ioread8(void *addr)
+{
+    uint8_t val;
+
+    val = (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint8_t *)addr :
+        inb((unsigned long)addr);
+
+    return val;
+}
+
+static inline uint16_t ioread16(void *addr)
+{
+    uint16_t val;
+
+    val = (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint16_t *)addr :
+        inw((unsigned long)addr);
+
+    return val;
+}
+
+static inline uint32_t ioread32(void *addr)
+{
+    uint32_t val;
+
+    val = (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint32_t *)addr :
+        inl((unsigned long)addr);
+
+    return val;
+}
+
+static inline void iowrite8(uint8_t val, void *addr)
+{
+    (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint8_t *)addr = val :
+        outb(val, (unsigned long)addr);
+}
+
+static inline void iowrite16(uint16_t val, void *addr)
+{
+    (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint16_t *)addr = val :
+        outw(val, (unsigned long)addr);
+}
+
+static inline void iowrite32(uint32_t val, void *addr)
+{
+    (uint64_t)(uintptr_t)addr >= PIO_MAX ?
+        *(volatile uint32_t *)addr = val :
+        outl(val, (unsigned long)addr);
+}
+
  void
  pci_uio_ioport_read(struct rte_pci_ioport *p,
              void *data, size_t len, off_t offset)
@@ -499,25 +581,13 @@
      for (d = data; len > 0; d += size, reg += size, len -= size) {
          if (len >= 4) {
              size = 4;
-#if defined(RTE_ARCH_X86)
-            *(uint32_t *)d = inl(reg);
-#else
-            *(uint32_t *)d = *(volatile uint32_t *)reg;
-#endif
+            *(uint32_t *)d = ioread32((void *)reg);
          } else if (len >= 2) {
              size = 2;
-#if defined(RTE_ARCH_X86)
-            *(uint16_t *)d = inw(reg);
-#else
-            *(uint16_t *)d = *(volatile uint16_t *)reg;
-#endif
+            *(uint16_t *)d = ioread16((void *)reg);
          } else {
              size = 1;
-#if defined(RTE_ARCH_X86)
-            *d = inb(reg);
-#else
-            *d = *(volatile uint8_t *)reg;
-#endif
+            *d = ioread8((void *)reg);
          }
      }
  }
@@ -533,25 +603,13 @@
      for (s = data; len > 0; s += size, reg += size, len -= size) {
          if (len >= 4) {
              size = 4;
-#if defined(RTE_ARCH_X86)
-            outl_p(*(const uint32_t *)s, reg);
-#else
-            *(volatile uint32_t *)reg = *(const uint32_t *)s;
-#endif
+            iowrite32(*(const uint32_t *)s, (void *)reg);
          } else if (len >= 2) {
              size = 2;
-#if defined(RTE_ARCH_X86)
-            outw_p(*(const uint16_t *)s, reg);
-#else
-            *(volatile uint16_t *)reg = *(const uint16_t *)s;
-#endif
+            iowrite16(*(const uint16_t *)s, (void *)reg);
          } else {
              size = 1;
-#if defined(RTE_ARCH_X86)
-            outb_p(*s, reg);
-#else
-            *(volatile uint8_t *)reg = *s;
-#endif
+            iowrite8(*s, (void *)reg);
          }
      }
  }
-- 
1.8.3.1


             reply	other threads:[~2020-09-14 12:44 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-09-14 12:43 谢华伟(此时此刻) [this message]
2020-09-14 13:09 谢华伟(此时此刻)
2020-09-15  8:18 [dpdk-dev] [PATCH] pci: support both PIO and MMIO BAR " 谢华伟(此时此刻)
2020-09-15 15:21 ` 谢华伟(此时此刻)
2020-09-24 18:02   ` Ferruh Yigit
2020-09-27 12:03     ` 谢华伟(此时此刻)

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=598a88dc-37c7-b525-9ab7-21f33415bfa2@alibaba-inc.com \
    --to=huawei.xhw@alibaba-inc.com \
    --cc=chenbo.xia@intel.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=zhihong.wang@intel.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).