[PATCH v6 1/5] server: Allow volume information queries to be asynchronous
Overall, this patchset has two objectives: 1) remove the volume information duplication in dlls/kernelbase/volume.c in preference to the version in dlls/mountmgr.sys/device.c 2) set the stage for fixing the advertisement of reparse point support ( https://github.com/wine-staging/wine-staging/blob/master/patches/ntdll-Junct... ) Per Zeb's suggestion, these patches are now organized as follows: 0001: allow volume information queries to be asynchronous 0002: allow NtQueryVolumeInformationFile to make async volume information queries 0003: implement volume information queries for device files 0004: hook up volume information queries for mountmgr 0005: reimplement GetVolumeInformation on top of GetVolumeInformationByHandle === This specific patch allows the volume information queries to be asynchronous on the server side. v6: No change v5: No change v4: Split content from server/device.c and dlls/ntoskrnl.exe/ntoskrnl.c and moved to patch 4 v3: No change v2: Split Best, Erich
Sorry I didn't notice these before, but... On 2/6/21 12:24 PM, Erich E. Hoover wrote:
From: "Erich E. Hoover" <erich.e.hoover(a)gmail.com> Subject: [PATCH v6 1/5] server: Allow volume information queries to be asynchronous Message-Id: <CAEU2+vqETWab9BeFry-kO9+9=+6Xvm47Azhi7jumnf3D8ej0Jg(a)mail.gmail.com> Date: Sat, 6 Feb 2021 11:24:44 -0700
Overall, this patchset has two objectives: 1) remove the volume information duplication in dlls/kernelbase/volume.c in preference to the version in dlls/mountmgr.sys/device.c 2) set the stage for fixing the advertisement of reparse point support ( https://github.com/wine-staging/wine-staging/blob/master/patches/ntdll-Junct... )
Per Zeb's suggestion, these patches are now organized as follows: 0001: allow volume information queries to be asynchronous 0002: allow NtQueryVolumeInformationFile to make async volume information queries 0003: implement volume information queries for device files 0004: hook up volume information queries for mountmgr 0005: reimplement GetVolumeInformation on top of GetVolumeInformationByHandle
===
This specific patch allows the volume information queries to be asynchronous on the server side.
v6: No change v5: No change v4: Split content from server/device.c and dlls/ntoskrnl.exe/ntoskrnl.c and moved to patch 4 v3: No change v2: Split
Best, Erich
From 01708bb64a4b20c8deba0b0a86b7e3a274090bd4 Mon Sep 17 00:00:00 2001 From: "Erich E. Hoover" <erich.e.hoover(a)gmail.com> Date: Sat, 23 May 2020 21:39:41 -0600 Subject: server: Allow volume information queries to be asynchronous.
Signed-off-by: Erich E. Hoover <erich.e.hoover(a)gmail.com> --- server/fd.c | 15 ++++++++++----- server/file.h | 4 ++-- server/named_pipe.c | 5 +++-- server/protocol.def | 2 ++ 4 files changed, 17 insertions(+), 9 deletions(-)
diff --git a/server/fd.c b/server/fd.c index 848203f31ec..ee3cb458d35 100644 --- a/server/fd.c +++ b/server/fd.c @@ -2314,9 +2314,10 @@ void default_fd_get_file_info( struct fd *fd, obj_handle_t handle, unsigned int }
/* default get_volume_info() routine */ -void no_fd_get_volume_info( struct fd *fd, unsigned int info_class ) +int no_fd_get_volume_info( struct fd *fd, struct async *async, unsigned int info_class ) { set_error( STATUS_OBJECT_TYPE_MISMATCH ); + return 0; }
/* default ioctl() routine */ @@ -2608,13 +2609,17 @@ DECL_HANDLER(get_file_info) /* query volume info */ DECL_HANDLER(get_volume_info) { - struct fd *fd = get_handle_fd_obj( current->process, req->handle, 0 ); + struct fd *fd = get_handle_fd_obj( current->process, req->async.handle, 0 ); + struct async *async;
- if (fd) + if (!fd) return; + + if ((async = create_request_async( fd, fd->comp_flags, &req->async ))) { - fd->fd_ops->get_volume_info( fd, req->info_class ); - release_object( fd ); + reply->wait = async_handoff( async, fd->fd_ops->get_volume_info( fd, async, req->info_class ), NULL, 0 );
Shouldn't we pass 1 to force_blocking here? From the client's perspective I believe this call is always synchronous. (Also, weird spacing before = in that line...)
+ release_object( async ); } + release_object( fd ); }
/* open a file object */ diff --git a/server/file.h b/server/file.h index 686bae084c5..a341793d4ab 100644 --- a/server/file.h +++ b/server/file.h @@ -65,7 +65,7 @@ struct fd_ops /* query file info */ void (*get_file_info)( struct fd *, obj_handle_t, unsigned int ); /* query volume info */ - void (*get_volume_info)( struct fd *, unsigned int ); + int (*get_volume_info)( struct fd *, struct async *, unsigned int ); /* perform an ioctl on the file */ int (*ioctl)(struct fd *fd, ioctl_code_t code, struct async *async ); /* queue an async operation */ @@ -112,7 +112,7 @@ extern int no_fd_write( struct fd *fd, struct async *async, file_pos_t pos ); extern int no_fd_flush( struct fd *fd, struct async *async ); extern void no_fd_get_file_info( struct fd *fd, obj_handle_t handle, unsigned int info_class ); extern void default_fd_get_file_info( struct fd *fd, obj_handle_t handle, unsigned int info_class ); -extern void no_fd_get_volume_info( struct fd *fd, unsigned int info_class ); +extern int no_fd_get_volume_info( struct fd *fd, struct async *async, unsigned int info_class ); extern int no_fd_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ); extern int default_fd_ioctl( struct fd *fd, ioctl_code_t code, struct async *async ); extern void no_fd_queue_async( struct fd *fd, struct async *async, int type, int count ); diff --git a/server/named_pipe.c b/server/named_pipe.c index 2e80ef3b1aa..a3ce9d463f1 100644 --- a/server/named_pipe.c +++ b/server/named_pipe.c @@ -146,7 +146,7 @@ static WCHAR *pipe_end_get_full_name( struct object *obj, data_size_t *len ); static int pipe_end_read( struct fd *fd, struct async *async, file_pos_t pos ); static int pipe_end_write( struct fd *fd, struct async *async_data, file_pos_t pos ); static int pipe_end_flush( struct fd *fd, struct async *async ); -static void pipe_end_get_volume_info( struct fd *fd, unsigned int info_class ); +static int pipe_end_get_volume_info( struct fd *fd, struct async *async, unsigned int info_class ); static void pipe_end_reselect_async( struct fd *fd, struct async_queue *queue ); static void pipe_end_get_file_info( struct fd *fd, obj_handle_t handle, unsigned int info_class );
@@ -742,7 +742,7 @@ static WCHAR *pipe_end_get_full_name( struct object *obj, data_size_t *len ) return pipe_end->pipe->obj.ops->get_full_name( &pipe_end->pipe->obj, len ); }
-static void pipe_end_get_volume_info( struct fd *fd, unsigned int info_class ) +static int pipe_end_get_volume_info( struct fd *fd, struct async *async, unsigned int info_class ) { switch (info_class) { @@ -762,6 +762,7 @@ static void pipe_end_get_volume_info( struct fd *fd, unsigned int info_class ) default: set_error( STATUS_NOT_IMPLEMENTED ); } + return 0; }
static void message_queue_read( struct pipe_end *pipe_end, struct iosb *iosb ) diff --git a/server/protocol.def b/server/protocol.def index 3747a3b63ec..cd7ed35d9a3 100644 --- a/server/protocol.def +++ b/server/protocol.def @@ -1400,9 +1400,11 @@ enum server_fd_type
/* Query volume information */ @REQ(get_volume_info) + async_data_t async; /* async I/O parameters */
Another inconsistent spacing nitpick here.
obj_handle_t handle; /* handle to the file */ unsigned int info_class; /* queried information class */ @REPLY + obj_handle_t wait; /* handle to wait on for blocking read */ VARARG(data,bytes); /* volume info data */ @END
-- 2.17.1
On Wed, Feb 10, 2021 at 9:15 PM Zebediah Figura (she/her) <zfigura(a)codeweavers.com> wrote:
Sorry I didn't notice these before, but...
That's alright, I always appreciate your feedback - it increases the chances that AJ will be happy ;) Best, Erich
On 2/6/21 12:24 PM, Erich E. Hoover wrote:
From: "Erich E. Hoover" <erich.e.hoover(a)gmail.com> Subject: [PATCH v6 1/5] server: Allow volume information queries to be asynchronous Message-Id: <CAEU2+vqETWab9BeFry-kO9+9=+6Xvm47Azhi7jumnf3D8ej0Jg(a)mail.gmail.com> Date: Sat, 6 Feb 2021 11:24:44 -0700 ... + if ((async = create_request_async( fd, fd->comp_flags, &req->async ))) { - fd->fd_ops->get_volume_info( fd, req->info_class ); - release_object( fd ); + reply->wait = async_handoff( async, fd->fd_ops->get_volume_info( fd, async, req->info_class ), NULL, 0 );
Shouldn't we pass 1 to force_blocking here? From the client's perspective I believe this call is always synchronous.
That makes sense to me, I'll go ahead and change it.
(Also, weird spacing before = in that line...)
Sorry about that, that's from copy-pasting from another instance of async_handoff and forgetting to delete the alignment spaces.
... static void message_queue_read( struct pipe_end *pipe_end, struct iosb *iosb ) diff --git a/server/protocol.def b/server/protocol.def index 3747a3b63ec..cd7ed35d9a3 100644 --- a/server/protocol.def +++ b/server/protocol.def @@ -1400,9 +1400,11 @@ enum server_fd_type
/* Query volume information */ @REQ(get_volume_info) + async_data_t async; /* async I/O parameters */
Another inconsistent spacing nitpick here.
Consider it taken care of.
...
participants (2)
-
Erich E. Hoover -
Zebediah Figura (she/her)