Since ib_copy_from_udata() doesn't check yet the available input data
length before accessing userspace memory, an explicit check of this
length is required to prevent:
- reading past the user provided buffer,
- underflow when subtracting the expected command size from the input
length.
This will ensure the newly added flow steering uverbs don't try to
process truncated commands.
Link: http://marc.info/?i=cover.1386798254.git.ydroneaud@opteya.com>
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
If the flow_spec items parsed count does not match the number of items
declared in the flow_attr command, or if not all bytes are used for
flow_spec items (eg. trailing garbage), a log message is reported and
the function leave through the error path. Unfortunately the error
code is currently not set.
This patch set error code to -EINVAL in such cases, so that the error
is reported to userspace instead of silently fail.
Link: http://marc.info/?i=cover.1386798254.git.ydroneaud@opteya.com>
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
As noted by Daniel Vetter in its article "Botching up ioctls"[1]
"Check *all* unused fields and flags and all the padding for whether
it's 0, and reject the ioctl if that's not the case. Otherwise
your nice plan for future extensions is going right down the
gutters since someone *will* submit an ioctl struct with random
stack garbage in the yet unused parts. Which then bakes in the ABI
that those fields can never be used for anything else but garbage."
It's important to ensure that reserved fields are set to known value,
so that it will be possible to use them latter to extend the ABI.
The same reasonning apply to comp_mask field present in newer uverbs
command: per commit 22878dbc91 ("IB/core: Better checking of
userspace values for receive flow steering"), unsupported values in
comp_mask are rejected.
[1] http://blog.ffwll.ch/2013/11/botching-up-ioctls.html
Link: http://marc.info/?i=cover.1386798254.git.ydroneaud@opteya.com>
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This commit reverts commit 7afbddfae9 ("IB/core: Temporarily disable
create_flow/destroy_flow uverbs"). Since the uverbs extensions
functionality was experimental for v3.12, this patch re-enables the
support for them and flow-steering for v3.13.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Commit 400dbc9658 ("IB/core: Infrastructure for extensible uverbs
commands") added an infrastructure for extensible uverbs commands
while later commit 436f2ad05a ("IB/core: Export ib_create/destroy_flow
through uverbs") exported ib_create_flow()/ib_destroy_flow() functions
using this new infrastructure.
According to the commit 400dbc9658, the purpose of this
infrastructure is to support passing around provider (eg. hardware)
specific buffers when userspace issue commands to the kernel, so that
it would be possible to extend uverbs (eg. core) buffers independently
from the provider buffers.
But the new kernel command function prototypes were not modified to
take advantage of this extension. This issue was exposed by Roland
Dreier in a previous review[1].
So the following patch is an attempt to a revised extensible command
infrastructure.
This improved extensible command infrastructure distinguish between
core (eg. legacy)'s command/response buffers from provider
(eg. hardware)'s command/response buffers: each extended command
implementing function is given a struct ib_udata to hold core
(eg. uverbs) input and output buffers, and another struct ib_udata to
hold the hw (eg. provider) input and output buffers.
Having those buffers identified separately make it easier to increase
one buffer to support extension without having to add some code to
guess the exact size of each command/response parts: This should make
the extended functions more reliable.
Additionally, instead of relying on command identifier being greater
than IB_USER_VERBS_CMD_THRESHOLD, the proposed infrastructure rely on
unused bits in command field: on the 32 bits provided by command
field, only 6 bits are really needed to encode the identifier of
commands currently supported by the kernel. (Even using only 6 bits
leaves room for about 23 new commands).
So this patch makes use of some high order bits in command field to
store flags, leaving enough room for more command identifiers than one
will ever need (eg. 256).
The new flags are used to specify if the command should be processed
as an extended one or a legacy one. While designing the new command
format, care was taken to make usage of flags itself extensible.
Using high order bits of the commands field ensure that newer
libibverbs on older kernel will properly fail when trying to call
extended commands. On the other hand, older libibverbs on newer kernel
will never be able to issue calls to extended commands.
The extended command header includes the optional response pointer so
that output buffer length and output buffer pointer are located
together in the command, allowing proper parameters checking. This
should make implementing functions easier and safer.
Additionally the extended header ensure 64bits alignment, while making
all sizes multiple of 8 bytes, extending the maximum buffer size:
legacy extended
Maximum command buffer: 256KBytes 1024KBytes (512KBytes + 512KBytes)
Maximum response buffer: 256KBytes 1024KBytes (512KBytes + 512KBytes)
For the purpose of doing proper buffer size accounting, the headers
size are no more taken in account in "in_words".
One of the odds of the current extensible infrastructure, reading
twice the "legacy" command header, is fixed by removing the "legacy"
command header from the extended command header: they are processed as
two different parts of the command: memory is read once and
information are not duplicated: it's making clear that's an extended
command scheme and not a different command scheme.
The proposed scheme will format input (command) and output (response)
buffers this way:
- command:
legacy header +
extended header +
command data (core + hw):
+----------------------------------------+
| flags | 00 00 | command |
| in_words | out_words |
+----------------------------------------+
| response |
| response |
| provider_in_words | provider_out_words |
| padding |
+----------------------------------------+
| |
. <uverbs input> .
. (in_words * 8) .
| |
+----------------------------------------+
| |
. <provider input> .
. (provider_in_words * 8) .
| |
+----------------------------------------+
- response, if present:
+----------------------------------------+
| |
. <uverbs output space> .
. (out_words * 8) .
| |
+----------------------------------------+
| |
. <provider output space> .
. (provider_out_words * 8) .
| |
+----------------------------------------+
The overall design is to ensure that the extensible infrastructure is
itself extensible while begin more reliable with more input and bound
checking.
Note:
The unused field in the extended header would be perfect candidate to
hold the command "comp_mask" (eg. bit field used to handle
compatibility). This was suggested by Roland Dreier in a previous
review[2]. But "comp_mask" field is likely to be present in the uverb
input and/or provider input, likewise for the response, as noted by
Matan Barak[3], so it doesn't make sense to put "comp_mask" in the
header.
[1]:
http://marc.info/?i=CAL1RGDWxmM17W2o_era24A-TTDeKyoL6u3NRu_=t_dhV_ZA9MA@mail.gmail.com
[2]:
http://marc.info/?i=CAL1RGDXJtrc849M6_XNZT5xO1+ybKtLWGq6yg6LhoSsKpsmkYA@mail.gmail.com
[3]:
http://marc.info/?i=525C1149.6000701@mellanox.com
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://marc.info/?i=cover.1383773832.git.ydroneaud@opteya.com
[ Convert "ret ? ret : 0" to the equivalent "ret". - Roland ]
Signed-off-by: Roland Dreier <roland@purestorage.com>
This patch adds "flow" prefix to most of data structure added as part
of commit 436f2ad05a ("IB/core: Export ib_create/destroy_flow through
uverbs") to keep those names in sync with the data structures added in
commit 319a441d13 ("IB/core: Add receive flow steering support").
It's just a matter of translating 'ib_flow' to 'ib_uverbs_flow'.
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://marc.info/?i=cover.1383773832.git.ydroneaud@opteya.com
Signed-off-by: Roland Dreier <roland@purestorage.com>
Commit 436f2ad05a ("IB/core: Export ib_create/destroy_flow through
uverbs") added public data structures to support receive flow
steering. The new structs are not following the 'uverbs' pattern:
they're lacking the common prefix 'ib_uverbs'.
This patch replaces ib_kern prefix by ib_uverbs.
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://marc.info/?i=cover.1383773832.git.ydroneaud@opteya.com
Signed-off-by: Roland Dreier <roland@purestorage.com>
This patch fixes the following issues:
1. Unneeded checks were removed
2. Removed the fixed size out of flow_attr.size, thus simplifying the checks.
3. Remove a 32bit hole on 64bit systems with strict alignment in
struct ib_kern_flow_att by adding a reserved field.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Enforce the rule that when requesting remote write or atomic permissions, local
write must be indicated as well. See IB spec 11.2.8.2.
Spotted by: Hagay Abramovsky <hagaya@mellanox.com>
Signed-off-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Currently, we don't copy the immediate data from the userspace struct
to the kernel one when UD messages are being sent.
This patch makes sure that the immediate data is set correctly.
Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The create_flow/destroy_flow uverbs and the associated extensions to
the user-kernel verbs ABI are under review and are too experimental to
freeze at this point.
So userspace is not exposed to experimental features and an uinstable
ABI, temporarily disable this for v3.12 (with a Kconfig option behind
staging to reenable it if desired).
The feature will be enabled after proper cleanup for v3.13.
Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://marc.info/?i=cover.1381351016.git.ydroneaud@opteya.com
Link: http://marc.info/?i=cover.1381177342.git.ydroneaud@opteya.com
[ Add a Kconfig option to reenable these verbs. - Roland ]
Signed-off-by: Roland Dreier <roland@purestorage.com>
- Don't allow unsupported comp_mask values, user should check
ibv_query_device to know which features are supported.
- Add a check in ib_uverbs_create_flow() to verify the size passed
from the user space.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Implement ib_uverbs_create_flow() and ib_uverbs_destroy_flow() to
support flow steering for user space applications.
Signed-off-by: Hadar Hen Zion <hadarh@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Added reference counting mechanism for XRC target QPs between
ib_uqp_object and its ib_uxrcd_object. This prevents closing an XRC
domain that is still attached to a QP. In addition, add missing code
in ib_uverbs_destroy_srq() to handle ib_uxrcd_object reference
counting correctly when destroying an xsrq.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
The macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe": O_CLOEXEC must
be used by default to not leak file descriptor across exec().
Replace calls to get_unused_fd() in uverbs with calls to
get_unused_fd_flags(O_CLOEXEC). Inheriting uverbs fds across exec()
cannot be used to do anything useful.
Based on a patch/suggestion from Yann Droneaud <ydroneaud@opteya.com>.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Convert to the much saner new idr interface.
v2: Mike triggered WARN_ON() in idr_preload() because send_mad(),
which may be used from non-process context, was calling
idr_preload() unconditionally. Preload iff @gfp_mask has
__GFP_WAIT.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Sean Hefty <sean.hefty@intel.com>
Reported-by: "Marciniszyn, Mike" <mike.marciniszyn@intel.com>
Cc: Roland Dreier <roland@kernel.org>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hal Rosenstock <hal.rosenstock@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs pile (part one) from Al Viro:
"Assorted stuff - cleaning namei.c up a bit, fixing ->d_name/->d_parent
locking violations, etc.
The most visible changes here are death of FS_REVAL_DOT (replaced with
"has ->d_weak_revalidate()") and a new helper getting from struct file
to inode. Some bits of preparation to xattr method interface changes.
Misc patches by various people sent this cycle *and* ocfs2 fixes from
several cycles ago that should've been upstream right then.
PS: the next vfs pile will be xattr stuff."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (46 commits)
saner proc_get_inode() calling conventions
proc: avoid extra pde_put() in proc_fill_super()
fs: change return values from -EACCES to -EPERM
fs/exec.c: make bprm_mm_init() static
ocfs2/dlm: use GFP_ATOMIC inside a spin_lock
ocfs2: fix possible use-after-free with AIO
ocfs2: Fix oops in ocfs2_fast_symlink_readpage() code path
get_empty_filp()/alloc_file() leave both ->f_pos and ->f_version zero
target: writev() on single-element vector is pointless
export kernel_write(), convert open-coded instances
fs: encode_fh: return FILEID_INVALID if invalid fid_type
kill f_vfsmnt
vfs: kill FS_REVAL_DOT by adding a d_weak_revalidate dentry op
nfsd: handle vfs_getattr errors in acl protocol
switch vfs_getattr() to struct path
default SET_PERSONALITY() in linux/elf.h
ceph: prepopulate inodes only when request is aborted
d_hash_and_lookup(): export, switch open-coded instances
9p: switch v9fs_set_create_acl() to inode+fid, do it before d_instantiate()
9p: split dropping the acls from v9fs_set_create_acl()
...
The existing user/kernel uverbs API has IB_USER_VERBS_CMD_ALLOC/DEALLOC_MW.
Implement these calls, along with destroying user memory windows during
process cleanup.
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Shani Michaeli <shanim@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
IB_QPT_RAW_PACKET allows applications to build a complete packet,
including L2 headers, when sending; on the receive side, the HW will
not strip any headers.
This QP type is designed for userspace direct access to Ethernet; for
example by applications that do TCP/IP themselves. Only processes
with the NET_RAW capability are allowed to create raw packet QPs (the
name "raw packet QP" is supposed to suggest an analogy to AF_PACKET /
SOL_RAW sockets).
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Add names for our lockdep classes, so instead of having to decipher
lockdep output with mysterious names:
Chain exists of:
key#14 --> key#11 --> key#13
lockdep will give us something nicer:
Chain exists of:
SRQ-uobj --> PD-uobj --> CQ-uobj
Signed-off-by: Roland Dreier <roland@purestorage.com>
We have just been investigating kernel panics related to
cq->ibcq.event_handler() completion calls. The problem is that
ib_destroy_qp() fails with -EBUSY.
Further investigation revealed qp->usecnt is not initialized. This
counter was introduced in linux-3.2 by commit 0e0ec7e063
("RDMA/core: Export ib_open_qp() to share XRC TGT QPs") but it only
gets initialized for IB_QPT_XRC_TGT, but it is checked in
ib_destroy_qp() for any QP type.
Fix this by initializing qp->usecnt for every QP we create.
Signed-off-by: Bernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Signed-off-by: Sven Breuner <sven.breuner@itwm.fraunhofer.de>
[ Initialize qp->usecnt in uverbs too. - Sean ]
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Clean up sparse warnings in the rdma core layer.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Userspace verbs multicast attach/detach operations on a QP are done
while holding the rwsem of the QP for reading. That's not sufficient
since a reader lock allows more than one reader to acquire the
lock. However, multicast attach/detach does list manipulation that
can corrupt the list if multiple threads run in parallel.
Fix this by acquiring the rwsem as a writer to serialize attach/detach
operations. Add idr_write_qp() and put_qp_write() to encapsulate
this.
This fixes oops seen when running applications that perform multicast
joins/leaves.
Reported by: Mike Dubman <miked@mellanox.com>
Signed-off-by: Eli Cohen <eli@mellanox.com>
Cc: <stable@kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow processes that share the same XRC domain to open an existing
shareable QP. This permits those processes to receive events on the
shared QP and transfer ownership, so that any process may modify the
QP. The latter allows the creating process to exit, while a remaining
process can still transition it for path migration purposes.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC TGT QPs are shared resources among multiple processes. Since the
creating process may exit, allow other processes which share the same
XRC domain to open an existing QP. This allows us to transfer
ownership of an XRC TGT QP to another process.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow user space to operate on XRC TGT QPs the same way as other types
of QPs, with one notable exception: since XRC TGT QPs may be shared
among multiple processes, the XRC TGT QP is allowed to exist beyond the
lifetime of the creating process.
The process that creates the QP is allowed to destroy it, but if the
process exits without destroying the QP, then the QP will be left bound
to the lifetime of the XRCD.
TGT QPs are not associated with CQs or a PD.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
XRC INI QPs are similar to send only RC QPs. Allow user space to create
INI QPs. Note that INI QPs do not require receive CQs.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
We require additional information to create XRC SRQs than we can
exchange using the existing create SRQ ABI. Provide an enhanced create
ABI for extended SRQ types.
Based on patches by Jack Morgenstein <jackm@dev.mellanox.co.il>
and Roland Dreier <roland@purestorage.com>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Allow user space to create XRC domains. Because XRCDs are expected to
be shared among multiple processes, we use inodes to identify an XRCD.
Based on patches by Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Currently, there is only a single ("basic") type of SRQ, but with XRC
support we will add a second. Prepare for this by defining an SRQ type
and setting all current users to IB_SRQT_BASIC.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
In ib_uverbs_poll_cq() code there is a potential integer overflow if
userspace passes in a large cmd.ne. The calls to kmalloc() would
allocate smaller buffers than intended, leading to memory corruption.
There iss also an information leak if resp wasn't all used.
Unprivileged userspace may call this function, although only if an
RDMA device that uses this function is present.
Fix this by copying CQ entries one at a time, which avoids the
allocation entirely, and also by moving this copying into a function
that makes sure to initialize all memory copied to userspace.
Special thanks to Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
for his help and advice.
Cc: <stable@kernel.org>
Signed-off-by: Dan Carpenter <error27@gmail.com>
[ Monkey around with things a bit to avoid bad code generation by gcc
when designated initializers are used. - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
This patch adds support for the IB "base memory management extension"
(BMME) and the equivalent iWARP operations (which the iWARP verbs
mandates all devices must implement). The new operations are:
- Allocate an ib_mr for use in fast register work requests.
- Allocate/free a physical buffer lists for use in fast register work
requests. This allows device drivers to allocate this memory as
needed for use in posting send requests (eg via dma_alloc_coherent).
- New send queue work requests:
* send with remote invalidate
* fast register memory region
* local invalidate memory region
* RDMA read with invalidate local memory region (iWARP only)
Consumer interface details:
- A new device capability flag IB_DEVICE_MEM_MGT_EXTENSIONS is added
to indicate device support for these features.
- New send work request opcodes IB_WR_FAST_REG_MR, IB_WR_LOCAL_INV,
IB_WR_RDMA_READ_WITH_INV are added.
- A new consumer API function, ib_alloc_mr() is added to allocate
fast register memory regions.
- New consumer API functions, ib_alloc_fast_reg_page_list() and
ib_free_fast_reg_page_list() are added to allocate and free
device-specific memory for fast registration page lists.
- A new consumer API function, ib_update_fast_reg_key(), is added to
allow the key portion of the R_Key and L_Key of a fast registration
MR to be updated. Consumers call this if desired before posting
a IB_WR_FAST_REG_MR work request.
Consumers can use this as follows:
- MR is allocated with ib_alloc_mr().
- Page list memory is allocated with ib_alloc_fast_reg_page_list().
- MR R_Key/L_Key "key" field is updated with ib_update_fast_reg_key().
- MR made VALID and bound to a specific page list via
ib_post_send(IB_WR_FAST_REG_MR)
- MR made INVALID via ib_post_send(IB_WR_LOCAL_INV),
ib_post_send(IB_WR_RDMA_READ_WITH_INV) or an incoming send with
invalidate operation.
- MR is deallocated with ib_dereg_mr()
- page lists dealloced via ib_free_fast_reg_page_list().
Applications can allocate a fast register MR once, and then can
repeatedly bind the MR to different physical block lists (PBLs) via
posting work requests to a send queue (SQ). For each outstanding
MR-to-PBL binding in the SQ pipe, a fast_reg_page_list needs to be
allocated (the fast_reg_page_list is owned by the low-level driver
from the consumer posting a work request until the request completes).
Thus pipelining can be achieved while still allowing device-specific
page_list processing.
The 32-bit fast register memory key/STag is composed of a 24-bit index
and an 8-bit key. The application can change the key each time it
fast registers thus allowing more control over the peer's use of the
key/STag (ie it can effectively be changed each time the rkey is
rebound to a page list).
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a new IB_WR_SEND_WITH_INV send opcode that can be used to mark a
"send with invalidate" work request as defined in the iWARP verbs and
the InfiniBand base memory management extensions. Also put "imm_data"
and a new "invalidate_rkey" member in a new "ex" union in struct
ib_send_wr. The invalidate_rkey member can be used to pass in an
R_Key/STag to be invalidated. Add this new union to struct
ib_uverbs_send_wr. Add code to copy the invalidate_rkey field in
ib_uverbs_post_send().
Fix up low-level drivers to deal with the change to struct ib_send_wr,
and just remove the imm_data initialization from net/sunrpc/xprtrdma/,
since that code never does any send with immediate operations.
Also, move the existing IB_DEVICE_SEND_W_INV flag to a new bit, since
the iWARP drivers currently in the tree set the bit. The amso1100
driver at least will silently fail to honor the IB_SEND_INVALIDATE bit
if passed in as part of userspace send requests (since it does not
implement kernel bypass work request queueing). Remove the flag from
all existing drivers that set it until we know which ones are OK.
The values chosen for the new flag is not consecutive to avoid clashing
with flags defined in the XRC patches, which are not merged yet but
which are already in use and are likely to be merged soon.
This resurrects a patch sent long ago by Mikkel Hagen <mhagen@iol.unh.edu>.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a create_flags member to struct ib_qp_init_attr that will allow a
kernel verbs consumer to create a pass special flags when creating a QP.
Add a flag value for telling low-level drivers that a QP will be used
for IPoIB UD LSO. The create_flags member will also be useful for XRC
and ehca low-latency QP support.
Since no create_flags handling is implemented yet, add code to all
low-level drivers to return -EINVAL if create_flags is non-zero.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Commit 9ead190b ("IB/uverbs: Don't serialize with ib_uverbs_idr_mutex")
rewrote how userspace objects are looked up in the uverbs module's
idrs, and introduced a severe bug in the process: there is no checking
that an operation is being performed by the right process any more.
Fix this by adding the missing check of uobj->context in __idr_get_uobj().
Apparently everyone is being very careful to only touch their own
objects, because this bug was introduced in June 2006 in 2.6.18, and
has gone undetected until now.
Cc: stable <stable@kernel.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Export ib_umem_get()/ib_umem_release() and put low-level drivers in
control of when to call ib_umem_get() to pin and DMA map userspace,
rather than always calling it in ib_uverbs_reg_mr() before calling the
low-level driver's reg_user_mr method.
Also move these functions to be in the ib_core module instead of
ib_uverbs, so that driver modules using them do not depend on
ib_uverbs.
This has a number of advantages:
- It is better design from the standpoint of making generic code a
library that can be used or overridden by device-specific code as
the details of specific devices dictate.
- Drivers that do not need to pin userspace memory regions do not
need to take the performance hit of calling ib_mem_get(). For
example, although I have not tried to implement it in this patch,
the ipath driver should be able to avoid pinning memory and just
use copy_{to,from}_user() to access userspace memory regions.
- Buffers that need special mapping treatment can be identified by
the low-level driver. For example, it may be possible to solve
some Altix-specific memory ordering issues with mthca CQs in
userspace by mapping CQ buffers with extra flags.
- Drivers that need to pin and DMA map userspace memory for things
other than memory regions can use ib_umem_get() directly, instead
of hacks using extra parameters to their reg_phys_mr method. For
example, the mlx4 driver that is pending being merged needs to pin
and DMA map QP and CQ buffers, but it does not need to create a
memory key for these buffers. So the cleanest solution is for mlx4
to call ib_umem_get() in the create_qp and create_cq methods.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Yann Droneaud
Roland Dreier
Matan Barak
Eli Cohen
Latchesar Ionkov
Hadar Hen Zion
Yishai Hadas
Tejun Heo
Linus Torvalds
Al Viro
Shani Michaeli
Or Gerlitz
Bernd Schubert
Sean Hefty
Dan Carpenter
Eli Cohen
Roel Kluin
Steve Wise
Roland Dreier