rpms/libusb1
7
0
Fork 0
Code Issues Pull Requests Releases Activity

Add some small error handling fixes

Browse Source
This commit is contained in:
Hans de Goede 2012-03-14 16:42:31 +01:00
parent 71e7ad084b
commit cd757587c6
5 changed files with 257 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
0001-linux-Fix-cancel_transfer-return-value-when-cancelli.patch Normal file
View File

@ -0,0 +1,37 @@
From 52508a86e26f0bc74b0a7a3b05ed08a29996b44c Mon Sep 17 00:00:00 2001
From: Hans de Goede <hdegoede@redhat.com>
Date: Mon, 20 Feb 2012 16:05:48 +0100
Subject: [PATCH 1/6] linux: Fix cancel_transfer return value when cancelling
a multi-urb transfer
If we fail to cancel the last urb of a multi-urb transfer because it
has already completed (errno == EINVAL on DISCARD_URB), then the entire
transfer has already completed, so returning NOT_FOUND is consistent with what
the documentation for libusb_cancel_transfer says.
But if we've successfully cancelled the last urb, and then another urb
fails with errno == EINVAL, this means that we've still cancelled the
transfer, as it has only *partially* completed.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
libusb/os/linux_usbfs.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/libusb/os/linux_usbfs.c b/libusb/os/linux_usbfs.c
index 2b81189..099fc61 100644
--- a/libusb/os/linux_usbfs.c
+++ b/libusb/os/linux_usbfs.c
@@ -1466,7 +1466,8 @@ static int discard_urbs(struct usbi_transfer *itransfer, int first, int last_plu
if (EINVAL == errno) {
usbi_dbg("URB not found --> assuming ready to be reaped");
- ret = LIBUSB_ERROR_NOT_FOUND;
+ if (i == (last_plus_one - 1))
+ ret = LIBUSB_ERROR_NOT_FOUND;
} else if (ENODEV == errno) {
usbi_dbg("Device not found for URB --> assuming ready to be reaped");
ret = LIBUSB_ERROR_NO_DEVICE;
--
1.7.9.3

46
0002-Don-t-print-errors-when-cancel_transfer-fails-with-N.patch Normal file
View File

@ -0,0 +1,46 @@
From e8c0b72bf8cc6d89c3546bbdbcc85b2c63086578 Mon Sep 17 00:00:00 2001
From: Hans de Goede <hdegoede@redhat.com>
Date: Mon, 20 Feb 2012 16:12:19 +0100
Subject: [PATCH 2/6] Don't print errors when cancel_transfer fails with
NOT_FOUND
As stated in the documentation for libusb_cancel_transfer,
LIBUSB_ERROR_NOT_FOUND is an expected return value for
libusb_cancel_transfer (under certain circumstances) printing
an error each time this happens therefor is undesirable.
More so because under Linux IOCTL_USBFS_DISCARDURB sets errno
to EINVAL when the kernel could not find the urb in the kernels
urbs in flight list. Which means that the urb has already completed
at the host controller level, but it has not necessarily already
been reaped. IOW under Linux libusb_cancel_transfer may yield a
result of LIBUSB_ERROR_NOT_FOUND *before* the transfer's callback
has been called! So there is no way for an application to avoid
calling libusb_cancel_transfer on already completed transfers.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
libusb/io.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/libusb/io.c b/libusb/io.c
index bb6e275..9f46cf0 100644
--- a/libusb/io.c
+++ b/libusb/io.c
@@ -1351,8 +1351,11 @@ int API_EXPORTED libusb_cancel_transfer(struct libusb_transfer *transfer)
usbi_mutex_lock(&itransfer->lock);
r = usbi_backend->cancel_transfer(itransfer);
if (r < 0) {
- usbi_err(TRANSFER_CTX(transfer),
- "cancel transfer failed error %d", r);
+ if (r != LIBUSB_ERROR_NOT_FOUND)
+ usbi_err(TRANSFER_CTX(transfer),
+ "cancel transfer failed error %d", r);
+ else
+ usbi_dbg("cancel transfer failed error %d", r);
if (r == LIBUSB_ERROR_NO_DEVICE)
itransfer->flags |= USBI_TRANSFER_DEVICE_DISAPPEARED;
--
1.7.9.3

92
0003-linux-Fix-handling-of-urb-status-codes.patch Normal file
View File

@ -0,0 +1,92 @@
From 3b6ed16cd2f098dd3920853d20940b3560c20ece Mon Sep 17 00:00:00 2001
From: Hans de Goede <hdegoede@redhat.com>
Date: Fri, 24 Feb 2012 10:24:00 +0100
Subject: [PATCH 3/6] linux: Fix handling of urb status codes
During testing of my usbredir code I hit a case where EOVERFLOW was not handled
in handle_control_completion. Instead of just fixing this one case I've audited
(and fixed where necessary) all handle_foo_completion functions to know about
all errors documented in linux/Documentation/usb/error-codes.txt.
Note that for handle_iso_completion this patch actually removes the handling
of some codes, since these can never occur on an iso urb (they can only
occur on the iso packets included in the urb, see the next patch in this
series). Also in case an unknown status is encountered on an iso urb, this
patch actually sets the urb's status to ERROR, rather then leaving it at
completed.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
libusb/os/linux_usbfs.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/libusb/os/linux_usbfs.c b/libusb/os/linux_usbfs.c
index 099fc61..36d37a4 100644
--- a/libusb/os/linux_usbfs.c
+++ b/libusb/os/linux_usbfs.c
@@ -1952,6 +1952,7 @@ static int handle_bulk_completion(struct usbi_transfer *itransfer,
case -ENOENT: /* cancelled */
case -ECONNRESET:
break;
+ case -ENODEV:
case -ESHUTDOWN:
usbi_dbg("device removed");
tpriv->reap_status = LIBUSB_TRANSFER_NO_DEVICE;
@@ -1970,6 +1971,8 @@ static int handle_bulk_completion(struct usbi_transfer *itransfer,
case -ETIME:
case -EPROTO:
case -EILSEQ:
+ case -ECOMM:
+ case -ENOSR:
usbi_dbg("low level error %d", urb->status);
tpriv->reap_action = ERROR;
goto cancel_remaining;
@@ -2081,19 +2084,16 @@ static int handle_iso_completion(struct usbi_transfer *itransfer,
case 0:
break;
case -ENOENT: /* cancelled */
+ case -ECONNRESET:
break;
case -ESHUTDOWN:
usbi_dbg("device removed");
status = LIBUSB_TRANSFER_NO_DEVICE;
break;
- case -ETIME:
- case -EPROTO:
- case -EILSEQ:
- usbi_dbg("low-level USB error %d", urb->status);
- break;
default:
usbi_warn(TRANSFER_CTX(transfer),
"unrecognised urb status %d", urb->status);
+ status = LIBUSB_TRANSFER_ERROR;
break;
}
@@ -2139,6 +2139,7 @@ static int handle_control_completion(struct usbi_transfer *itransfer,
case -ENOENT: /* cancelled */
status = LIBUSB_TRANSFER_CANCELLED;
break;
+ case -ENODEV:
case -ESHUTDOWN:
usbi_dbg("device removed");
status = LIBUSB_TRANSFER_NO_DEVICE;
@@ -2147,9 +2148,15 @@ static int handle_control_completion(struct usbi_transfer *itransfer,
usbi_dbg("unsupported control request");
status = LIBUSB_TRANSFER_STALL;
break;
+ case -EOVERFLOW:
+ usbi_dbg("control overflow error");
+ status = LIBUSB_TRANSFER_OVERFLOW;
+ break;
case -ETIME:
case -EPROTO:
case -EILSEQ:
+ case -ECOMM:
+ case -ENOSR:
usbi_dbg("low-level bus error occurred");
status = LIBUSB_TRANSFER_ERROR;
break;
--
1.7.9.3

68
0004-linux-Translate-linux-iso-pkt-status-codes-to-libusb.patch Normal file
View File

@ -0,0 +1,68 @@
From 4c3e7f9818c0d1d0462fde6f219da65fb102a434 Mon Sep 17 00:00:00 2001
From: Hans de Goede <hdegoede@redhat.com>
Date: Fri, 24 Feb 2012 11:15:30 +0100
Subject: [PATCH 4/6] linux: Translate linux iso pkt status codes to libusb
transfer status codes
During testing of my usbredir code I hit a scenario where my libusb app
was seeing EXDEV as status in the transfer's iso_packet_desc
This happened because we don't translate linux negative errno errors
stored in iso pkts status to libusb transfer status codes at all! So this
patch adds translation for this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
---
libusb/os/linux_usbfs.c | 36 +++++++++++++++++++++++++++++++++++-
1 file changed, 35 insertions(+), 1 deletion(-)
diff --git a/libusb/os/linux_usbfs.c b/libusb/os/linux_usbfs.c
index 36d37a4..a7d8298 100644
--- a/libusb/os/linux_usbfs.c
+++ b/libusb/os/linux_usbfs.c
@@ -2053,7 +2053,41 @@ static int handle_iso_completion(struct usbi_transfer *itransfer,
struct usbfs_iso_packet_desc *urb_desc = &urb->iso_frame_desc[i];
struct libusb_iso_packet_descriptor *lib_desc =
&transfer->iso_packet_desc[tpriv->iso_packet_offset++];
- lib_desc->status = urb_desc->status;
+ lib_desc->status = LIBUSB_TRANSFER_COMPLETED;
+ switch (urb_desc->status) {
+ case 0:
+ break;
+ case -ENOENT: /* cancelled */
+ case -ECONNRESET:
+ break;
+ case -ENODEV:
+ case -ESHUTDOWN:
+ usbi_dbg("device removed");
+ lib_desc->status = LIBUSB_TRANSFER_NO_DEVICE;
+ break;
+ case -EPIPE:
+ usbi_dbg("detected endpoint stall");
+ lib_desc->status = LIBUSB_TRANSFER_STALL;
+ break;
+ case -EOVERFLOW:
+ usbi_dbg("overflow error");
+ lib_desc->status = LIBUSB_TRANSFER_OVERFLOW;
+ break;
+ case -ETIME:
+ case -EPROTO:
+ case -EILSEQ:
+ case -ECOMM:
+ case -ENOSR:
+ case -EXDEV:
+ usbi_dbg("low-level USB error %d", urb_desc->status);
+ lib_desc->status = LIBUSB_TRANSFER_ERROR;
+ break;
+ default:
+ usbi_warn(TRANSFER_CTX(transfer),
+ "unrecognised urb status %d", urb_desc->status);
+ lib_desc->status = LIBUSB_TRANSFER_ERROR;
+ break;
+ }
lib_desc->actual_length = urb_desc->actual_length;
}
--
1.7.9.3

15
libusb1.spec
View File

@ -1,7 +1,7 @@
Summary: A library which allows userspace access to USB devices
Name: libusb1
Version: 1.0.9
Release: 0.4.rc1%{?dist}
Release: 0.5.rc1%{?dist}
# This is a git snapshot of what will hopefully soon become 1.0.9, but
# we need this now, to get things in place for:
# http://fedoraproject.org/wiki/Features/UsbNetworkRedirection
@ -14,6 +14,12 @@ Release: 0.4.rc1%{?dist}
# mv libusb-1.0.8.tar.bz2 libusb-1.0.9-rc1.tar.bz2
Source0: libusb-1.0.9-rc1.tar.bz2
#Source0: http://downloads.sourceforge.net/libusb/libusb-%{version}.tar.bz2
Patch1: 0001-linux-Fix-cancel_transfer-return-value-when-cancelli.patch
Patch2: 0002-Don-t-print-errors-when-cancel_transfer-fails-with-N.patch
Patch3: 0003-linux-Fix-handling-of-urb-status-codes.patch
Patch4: 0004-linux-Translate-linux-iso-pkt-status-codes-to-libusb.patch
License: LGPLv2+
Group: System Environment/Libraries
URL: http://libusb.wiki.sourceforge.net/Libusb1.0
@ -54,6 +60,10 @@ This package contains static libraries to develop applications that use libusb1.
%prep
%setup -q -n libusb-1.0.8
%patch1 -p1
%patch2 -p1
%patch3 -p1
%patch4 -p1
%build
%configure --libdir=/%{_lib}
@ -97,6 +107,9 @@ mv %{buildroot}/%{_lib}/pkgconfig/* %{buildroot}%{_libdir}/pkgconfig/
/%{_lib}/*.a
%changelog
* Wed Mar 14 2012 Hans de Goede <hdegoede@redhat.com> - 1.0.9-0.5.rc1
- Add some small error handling fixes
* Fri Jan 13 2012 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 1.0.9-0.4.rc1
- Rebuilt for https://fedoraproject.org/wiki/Fedora_17_Mass_Rebuild