@@ -9,16 +9,117 @@
#include <linux/pci.h>
#include <linux/slab.h>
-#include <linux/anon_inodes.h>
#include <linux/file.h>
#include <misc/cxl.h>
-#include <linux/fs.h>
+#include <linux/module.h>
+#include <linux/mount.h>
#include "cxl.h"
+/*
+ * Since we want to track memory mappings to be able to force-unmap
+ * when the AFU is no longer reachable, we need an inode. For devices
+ * opened through the cxl user API, this is not a problem, but a
+ * userland process can also get a cxl fd through the cxl_get_fd()
+ * API, which is used by the cxlflash driver.
+ *
+ * Therefore we implement our own simple pseudo-filesystem and inode
+ * allocator. We don't use the anonymous inode, as we need the
+ * meta-data associated with it (address_space) and it is shared by
+ * other drivers/processes, so it could lead to cxl unmapping VMAs
+ * from random processes.
+ */
+
+#define CXL_PSEUDO_FS_MAGIC 0x1697697f
+
+static int cxl_fs_cnt;
+static struct vfsmount *cxl_vfs_mount;
+
+static const struct dentry_operations cxl_fs_dops = {
+ .d_dname = simple_dname,
+};
+
+static struct dentry *cxl_fs_mount(struct file_system_type *fs_type, int flags,
+ const char *dev_name, void *data)
+{
+ return mount_pseudo(fs_type, "cxl:", NULL, &cxl_fs_dops,
+ CXL_PSEUDO_FS_MAGIC);
+}
+
+static struct file_system_type cxl_fs_type = {
+ .name = "cxl",
+ .owner = THIS_MODULE,
+ .mount = cxl_fs_mount,
+ .kill_sb = kill_anon_super,
+};
+
+
+void cxl_release_mapping(struct cxl_context *ctx)
+{
+ if (ctx->kernelapi && ctx->mapping)
+ simple_release_fs(&cxl_vfs_mount, &cxl_fs_cnt);
+}
+
+static struct file *cxl_getfile(const char *name,
+ const struct file_operations *fops,
+ void *priv, int flags)
+{
+ struct qstr this;
+ struct path path;
+ struct file *file;
+ struct inode *inode = NULL;
+ int rc;
+
+ /* strongly inspired by anon_inode_getfile() */
+
+ if (fops->owner && !try_module_get(fops->owner))
+ return ERR_PTR(-ENOENT);
+
+ rc = simple_pin_fs(&cxl_fs_type, &cxl_vfs_mount, &cxl_fs_cnt);
+ if (rc < 0) {
+ pr_err("Cannot mount cxl pseudo filesystem: %d\n", rc);
+ file = ERR_PTR(rc);
+ goto err_module;
+ }
+
+ inode = alloc_anon_inode(cxl_vfs_mount->mnt_sb);
+ if (IS_ERR(inode)) {
+ file = ERR_CAST(inode);
+ goto err_fs;
+ }
+
+ file = ERR_PTR(-ENOMEM);
+ this.name = name;
+ this.len = strlen(name);
+ this.hash = 0;
+ path.dentry = d_alloc_pseudo(cxl_vfs_mount->mnt_sb, &this);
+ if (!path.dentry)
+ goto err_inode;
+
+ path.mnt = mntget(cxl_vfs_mount);
+ d_instantiate(path.dentry, inode);
+
+ file = alloc_file(&path, OPEN_FMODE(flags), fops);
+ if (IS_ERR(file))
+ goto err_dput;
+ file->f_flags = flags & (O_ACCMODE | O_NONBLOCK);
+ file->private_data = priv;
+
+ return file;
+
+err_dput:
+ path_put(&path);
+err_inode:
+ iput(inode);
+err_fs:
+ simple_release_fs(&cxl_vfs_mount, &cxl_fs_cnt);
+err_module:
+ module_put(fops->owner);
+ return file;
+}
+
struct cxl_context *cxl_dev_context_init(struct pci_dev *dev)
{
- struct address_space *mapping;
struct cxl_afu *afu;
struct cxl_context *ctx;
int rc;
@@ -33,28 +134,13 @@ struct cxl_context *cxl_dev_context_init(struct pci_dev *dev)
ctx->kernelapi = true;
- /*
- * Make our own address space since we won't have one from the
- * filesystem like the user api has, and even if we do associate a file
- * with this context we don't want to use the global anonymous inode's
- * address space as that can invalidate unrelated users:
- */
- mapping = kmalloc(sizeof(struct address_space), GFP_KERNEL);
- if (!mapping) {
- rc = -ENOMEM;
- goto err_ctx;
- }
- address_space_init_once(mapping);
-
/* Make it a slave context. We can promote it later? */
- rc = cxl_context_init(ctx, afu, false, mapping);
+ rc = cxl_context_init(ctx, afu, false);
if (rc)
- goto err_mapping;
+ goto err_ctx;
return ctx;
-err_mapping:
- kfree(mapping);
err_ctx:
kfree(ctx);
err_dev:
@@ -269,6 +355,11 @@ struct file *cxl_get_fd(struct cxl_context *ctx, struct file_operations *fops,
{
struct file *file;
int rc, flags, fdtmp;
+ char *name = NULL;
+
+ /* only allow one per context */
+ if (ctx->mapping)
+ return ERR_PTR(-EEXIST);
flags = O_RDWR | O_CLOEXEC;
@@ -292,12 +383,13 @@ struct file *cxl_get_fd(struct cxl_context *ctx, struct file_operations *fops,
} else /* use default ops */
fops = (struct file_operations *)&afu_fops;
- file = anon_inode_getfile("cxl", fops, ctx, flags);
+ name = kasprintf(GFP_KERNEL, "cxl:%d", ctx->pe);
+ file = cxl_getfile(name, fops, ctx, flags);
+ kfree(name);
if (IS_ERR(file))
goto err_fd;
- file->f_mapping = ctx->mapping;
-
+ cxl_context_set_mapping(ctx, file->f_mapping);
*fd = fdtmp;
return file;
@@ -34,8 +34,7 @@ struct cxl_context *cxl_context_alloc(void)
/*
* Initialises a CXL context.
*/
-int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master,
- struct address_space *mapping)
+int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master)
{
int i;
@@ -44,7 +43,7 @@ int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master,
ctx->master = master;
ctx->pid = ctx->glpid = NULL; /* Set in start work ioctl */
mutex_init(&ctx->mapping_lock);
- ctx->mapping = mapping;
+ ctx->mapping = NULL;
/*
* Allocate the segment table before we put it in the IDR so that we
@@ -111,6 +110,14 @@ int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master,
return 0;
}
+void cxl_context_set_mapping(struct cxl_context *ctx,
+ struct address_space *mapping)
+{
+ mutex_lock(&ctx->mapping_lock);
+ ctx->mapping = mapping;
+ mutex_unlock(&ctx->mapping_lock);
+}
+
static int cxl_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
struct cxl_context *ctx = vma->vm_file->private_data;
@@ -294,8 +301,6 @@ static void reclaim_ctx(struct rcu_head *rcu)
if (ctx->ff_page)
__free_page(ctx->ff_page);
ctx->sstp = NULL;
- if (ctx->kernelapi)
- kfree(ctx->mapping);
if (ctx->irq_bitmap)
kfree(ctx->irq_bitmap);
@@ -308,6 +313,8 @@ static void reclaim_ctx(struct rcu_head *rcu)
void cxl_context_free(struct cxl_context *ctx)
{
+ if (ctx->kernelapi && ctx->mapping)
+ cxl_release_mapping(ctx);
mutex_lock(&ctx->afu->contexts_lock);
idr_remove(&ctx->afu->contexts_idr, ctx->pe);
mutex_unlock(&ctx->afu->contexts_lock);
@@ -763,8 +763,9 @@ void cxl_dump_debug_buffer(void *addr, size_t size);
void init_cxl_native(void);
struct cxl_context *cxl_context_alloc(void);
-int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master,
- struct address_space *mapping);
+int cxl_context_init(struct cxl_context *ctx, struct cxl_afu *afu, bool master);
+void cxl_context_set_mapping(struct cxl_context *ctx,
+ struct address_space *mapping);
void cxl_context_free(struct cxl_context *ctx);
int cxl_context_iomap(struct cxl_context *ctx, struct vm_area_struct *vma);
unsigned int cxl_map_irq(struct cxl *adapter, irq_hw_number_t hwirq,
@@ -818,6 +819,7 @@ int cxl_psl_purge(struct cxl_afu *afu);
void cxl_stop_trace(struct cxl *cxl);
int cxl_pci_vphb_add(struct cxl_afu *afu);
void cxl_pci_vphb_remove(struct cxl_afu *afu);
+void cxl_release_mapping(struct cxl_context *ctx);
extern struct pci_driver cxl_pci_driver;
extern struct platform_driver cxl_of_driver;
@@ -86,9 +86,12 @@ static int __afu_open(struct inode *inode, struct file *file, bool master)
goto err_put_afu;
}
- if ((rc = cxl_context_init(ctx, afu, master, inode->i_mapping)))
+ rc = cxl_context_init(ctx, afu, master);
+ if (rc)
goto err_put_afu;
+ cxl_context_set_mapping(ctx, inode->i_mapping);
+
pr_devel("afu_open pe: %i\n", ctx->pe);
file->private_data = ctx;
cxl_ctx_get();