[PULL REQUEST] NVMe driver updates

[PULL REQUEST] NVMe driver updates

[Matthew Wilcox]

Hi Linus,

Lots of exciting new features in the NVM Express driver this time,
including support for emulating SCSI commands, discard support and the
ability to submit per-sector metadata with I/Os.  It's still mostly
bugfixes though!

The following changes since commit a12183c62717ac4579319189a00f5883a18dff08:

  Merge branch 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip (2013-03-25 18:03:34 -0700)

are available in the git repository at:


  git://git.infradead.org/users/willy/linux-nvme.git master

for you to fetch changes up to 94f370cab6e5ac514b658c6b2b3aa308cefc5c7a:

  NVMe: Use user defined admin ioctl timeout (2013-05-09 16:03:50 -0400)

----------------------------------------------------------------
Arjan van de Ven (2):
      NVMe: Use round_jiffies_relative() for the periodic, once-per-second timer
      NVMe: Set TASK_INTERRUPTIBLE before processing queues

Keith Busch (13):
      NVMe: Add discard support for capable devices
      NVMe: Add a character device for each nvme device
      NVMe: queue usage fixes in nvme-scsi
      NVMe: Add scsi unmap to SG_IO
      NVMe: Free admin queue on request_irq error
      NVMe: Fix error clean-up on nvme_alloc_queue
      NVMe: Check for NULL memory in nvme_dev_add
      NVMe: Remove dead code in nvme_dev_add
      NVMe: Split non-mergeable bio requests
      NVMe: Device specific stripe size handling
      NVMe: Meta-data support in NVME_IOCTL_SUBMIT_IO
      NVMe: Schedule timeout for sync commands
      NVMe: Use user defined admin ioctl timeout

Matthew Wilcox (7):
      NVMe: Abstract out sector to block number conversion
      NVMe: Don't fail initialisation unnecessarily
      NVMe: Fix I/O cancellation status on big-endian machines
      NVMe: Fix endian-related problems in user I/O submission path
      NVMe: Wait for device to acknowledge shutdown
      NVMe: Only clear the enable bit when disabling controller
      NVMe: Simplify Firmware Activate code slightly

Vishal Verma (5):
      NVMe: Rename nvme.c to nvme-core.c
      NVMe: Move structures & definitions to header file
      NVMe: Add definitions for format command
      NVMe: Add nvme-scsi.c
      NVMe: Fix sparse warnings in scsi emulation

 drivers/block/Makefile                |    1 +
 drivers/block/{nvme.c => nvme-core.c} |  594 +++++--
 drivers/block/nvme-scsi.c             | 3053 +++++++++++++++++++++++++++++++++
 include/linux/nvme.h                  |  158 +-
 4 files changed, 3640 insertions(+), 166 deletions(-)
 rename drivers/block/{nvme.c => nvme-core.c} (79%)
 create mode 100644 drivers/block/nvme-scsi.c

NVMe: Add nvme-scsi.c (was Re: [PULL REQUEST] NVMe driver updates)

[Paolo Bonzini]

Il 09/05/2013 22:20, Matthew Wilcox ha scritto:
>       NVMe: Add nvme-scsi.c

I couldn't find the original patch on LKML, so I'll just quote the
relevant piece of code.

> +int nvme_sg_io(struct nvme_ns *ns, struct sg_io_hdr __user *u_hdr)
> +{
> +	struct sg_io_hdr hdr;
> +	int retcode;
> +
> +	if (!capable(CAP_SYS_ADMIN))
> +		return -EACCES;

This should be EPERM, and also it should use the same checks as other
implementations of SG_IO, including:

- checking if it's operating on a partition, and requiring CAP_SYS_RAWIO
if so;

- allowing a limited number of commands even for !capable(CAP_SYS_RAWIO).

All this is done by scsi_cmd_blk_ioctl.

Paolo

> +	if (copy_from_user(&hdr, u_hdr, sizeof(hdr)))
> +		return -EFAULT;
> +	if (hdr.interface_id != 'S')
> +		return -EINVAL;
> +	if (hdr.cmd_len > BLK_MAX_CDB)
> +		return -EINVAL;
> +
> +	retcode = nvme_scsi_translate(ns, &hdr);
> +	if (retcode < 0)
> +		return retcode;
> +	if (retcode > 0)
> +		retcode = SNTI_TRANSLATION_SUCCESS;
> +	if (copy_to_user(__user u_hdr, &hdr, sizeof(sg_io_hdr_t)) > 0)
> +		return -EFAULT;
> +
> +	return retcode;
> +}