io_uring: let to set a range for file slot allocation

From recently io_uring provides an option to allocate a file index for
operation registering fixed files. However, it's utterly unusable with
mixed approaches when for a part of files the userspace knows better
where to place it, as it may race and users don't have any sane way to
pick a slot and hoping it will not be taken.

Let the userspace to register a range of fixed file slots in which the
auto-allocation happens. The use case is splittting the fixed table in
two parts, where on of them is used for auto-allocation and another for
slot-specified operations.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/66ab0394e436f38437cf7c44676e1920d09687ad.1656154403.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 3ca8f36..26ef11e 100644 (file)
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -233,6 +233,9 @@ struct io_ring_ctx {
 
        unsigned long           check_cq;
 
+       unsigned int            file_alloc_start;
+       unsigned int            file_alloc_end;
+
        struct {
                /*
                 * We cache a range of free CQEs we can use, once exhausted it

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index f378eab..cf95354 100644 (file)
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -449,6 +449,9 @@ enum {
        /* sync cancelation API */
        IORING_REGISTER_SYNC_CANCEL             = 24,
 
+       /* register a range of fixed file slots for automatic slot allocation */
+       IORING_REGISTER_FILE_ALLOC_RANGE        = 25,
+
        /* this goes last */
        IORING_REGISTER_LAST
 };
@@ -595,4 +598,14 @@ struct io_uring_sync_cancel_reg {
        __u64                           pad[4];
 };
 
+/*
+ * Argument for IORING_REGISTER_FILE_ALLOC_RANGE
+ * The range is specified as [off, off + len)
+ */
+struct io_uring_file_index_range {
+       __u32   off;
+       __u32   len;
+       __u64   resv;
+};
+
 #endif

diff --git a/io_uring/filetable.c b/io_uring/filetable.c
index abaa5ba..7b47325 100644 (file)
--- a/io_uring/filetable.c
+++ b/io_uring/filetable.c
@@ -16,7 +16,7 @@
 static int io_file_bitmap_get(struct io_ring_ctx *ctx)
 {
        struct io_file_table *table = &ctx->file_table;
-       unsigned long nr = ctx->nr_user_files;
+       unsigned long nr = ctx->file_alloc_end;
        int ret;
 
        do {
@@ -24,11 +24,10 @@ static int io_file_bitmap_get(struct io_ring_ctx *ctx)
                if (ret != nr)
                        return ret;
 
-               if (!table->alloc_hint)
+               if (table->alloc_hint == ctx->file_alloc_start)
                        break;
-
                nr = table->alloc_hint;
-               table->alloc_hint = 0;
+               table->alloc_hint = ctx->file_alloc_start;
        } while (1);
 
        return -ENFILE;
@@ -175,3 +174,20 @@ int io_fixed_fd_remove(struct io_ring_ctx *ctx, unsigned int offset)
        io_rsrc_node_switch(ctx, ctx->file_data);
        return 0;
 }
+
+int io_register_file_alloc_range(struct io_ring_ctx *ctx,
+                                struct io_uring_file_index_range __user *arg)
+{
+       struct io_uring_file_index_range range;
+       u32 end;
+
+       if (copy_from_user(&range, arg, sizeof(range)))
+               return -EFAULT;
+       if (check_add_overflow(range.off, range.len, &end))
+               return -EOVERFLOW;
+       if (range.resv || end > ctx->nr_user_files)
+               return -EINVAL;
+
+       io_file_table_set_alloc_range(ctx, range.off, range.len);
+       return 0;
+}

diff --git a/io_uring/filetable.h b/io_uring/filetable.h
index 79eb50c..ff3a712 100644 (file)
--- a/io_uring/filetable.h
+++ b/io_uring/filetable.h
@@ -3,9 +3,7 @@
 #define IOU_FILE_TABLE_H
 
 #include <linux/file.h>
-
-struct io_ring_ctx;
-struct io_kiocb;
+#include <linux/io_uring_types.h>
 
 /*
  * FFS_SCM is only available on 64-bit archs, for 32-bit we just define it as 0
@@ -33,6 +31,9 @@ int __io_fixed_fd_install(struct io_ring_ctx *ctx, struct file *file,
                                unsigned int file_slot);
 int io_fixed_fd_remove(struct io_ring_ctx *ctx, unsigned int offset);
 
+int io_register_file_alloc_range(struct io_ring_ctx *ctx,
+                                struct io_uring_file_index_range __user *arg);
+
 unsigned int io_file_get_flags(struct file *file);
 
 static inline void io_file_bitmap_clear(struct io_file_table *table, int bit)
@@ -71,4 +72,17 @@ static inline void io_fixed_file_set(struct io_fixed_file *file_slot,
        file_slot->file_ptr = file_ptr;
 }
 
+static inline void io_reset_alloc_hint(struct io_ring_ctx *ctx)
+{
+       ctx->file_table.alloc_hint = ctx->file_alloc_start;
+}
+
+static inline void io_file_table_set_alloc_range(struct io_ring_ctx *ctx,
+                                                unsigned off, unsigned len)
+{
+       ctx->file_alloc_start = off;
+       ctx->file_alloc_end = off + len;
+       io_reset_alloc_hint(ctx);
+}
+
 #endif

diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 070ee9e..7452649 100644 (file)
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3866,6 +3866,12 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
                        break;
                ret = io_sync_cancel(ctx, arg);
                break;
+       case IORING_REGISTER_FILE_ALLOC_RANGE:
+               ret = -EINVAL;
+               if (!arg || nr_args)
+                       break;
+               ret = io_register_file_alloc_range(ctx, arg);
+               break;
        default:
                ret = -EINVAL;
                break;

diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
index 706fa02..d2e589c 100644 (file)
--- a/io_uring/rsrc.c
+++ b/io_uring/rsrc.c
@@ -1012,6 +1012,8 @@ int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
                io_file_bitmap_set(&ctx->file_table, i);
        }
 
+       /* default it to the whole table */
+       io_file_table_set_alloc_range(ctx, 0, ctx->nr_user_files);
        io_rsrc_node_switch(ctx, NULL);
        return 0;
 fail: