- Rename these so they are easy to enable and search for as a set
- Move the tracepoints to get a more accurate sense of control flow
- Tracepoints should not fire on xprt shutdown
- Display memory address in case data structure had been corrupted
- Abandon dprintk in these paths

I haven't ever gotten one of these tracepoints to trigger. I wonder
if we should simply remove them.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Capture obvious events and replace dprintk() call sites. Introduce
infrastructure so that adding more tracepoints in this code later
is simplified.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Try to capture DRC failures.

Two additional clean-ups:
- Introduce Doxygen-style comments for the main entry points
- Remove a dprintk that fires for an allocation failure. This was
  the only dprintk in the REPCACHE class.
Reported-by: kbuild test robot <lkp@intel.com>
[ cel: force typecast for display of checksum values ]
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Add missing TRACE_DEFINE_ENUMs in
include/trace/events/sunrpc.h
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up. At this point, we are not ready yet to support bio_vecs in
the UDP transport implementation. However, we can clean up
svc_udp_recvfrom() to match the tracing and straight-lining recently
changes made in svc_tcp_recvfrom().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

This function is not currently "generic" so remove the documenting
comment and rename it appropriately. Its internals are converted to
use bio_vecs for reading from the transport socket.

In existing typical sunrpc uses of bio_vecs, the bio_vec array is
allocated dynamically. Here, instead, an array of bio_vecs is added
to svc_rqst. The lifetime of this array can be greater than one call
to xpo_recvfrom():

- Multiple calls to xpo_recvfrom() might be needed to read an RPC
  message completely.

- At some later point, rq_arg.bvecs will point to this array and it
  will carry the received message into svc_process().

I also expect that a future optimization will remove either the
rq_vec or rq_pages array in favor of rq_bvec, thus conserving the
size of struct svc_rqst.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Rename these functions using the convention used for other xpo
method entry points.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: move exception processing out of the main path.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Refactor: svc_recvfrom() is going to be converted to read into
bio_vecs in a moment. Unhook the only other caller,
svc_tcp_recv_record(), which just wants to read the 4-byte stream
record marker into a kvec.

While we're in the area, streamline this helper by straight-lining
the hot path, replace dprintk call sites with tracepoints, and
reduce the number of atomic bit operations in this path.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up. I find the name of the svc_sock::sk_reclen field
confusing, so I've changed it to better reflect its function. This
field is not read directly to get the record length. Rather, it is
a buffer containing a record marker that needs to be decoded.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Report TCP socket state changes and accept failures via
tracepoints, replacing dprintk() call sites.

No tracepoint is added in svc_tcp_listen_data_ready. There's
no information available there that isn't also reported by the
svcsock_new_socket and the accept failure tracepoints.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

In addition to tracing recently-updated socket sendto events, this
commit adds a trace event class that can be used for additional
svcsock-related tracepoints in subsequent patches.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Commit 850cbadd ("udp: use it's own memory accounting
schema") removed the last skb-related tracepoint from svcsock.c, so
it is no longer necessary to include trace/events/skb.h.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

In lieu of dprintks or tracepoints in each individual transport
implementation, introduce tracepoints in the generic part of the RPC
layer. These typically fire for connection lifetime events, so
shouldn't contribute a lot of noise.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Capture transport creation failures.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: The xprt=%p was meant to distinguish events from different
transports, but the addr=%s does that just as well and does not
expose kernel memory addresses.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Failure to accept a connection is typically due to a problem
specific to a transport type. Also, ->xpo_accept returns NULL
on error rather than reporting a specific problem.

So, add failure-specific tracepoints in svc_rdma_accept().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: After commit 1e091c3b ("svcrdma: Ignore source port
when computing DRC hash"), the IP address stored in xpt_remote
always has a port number of zero. Thus, there's no need to display
the port number when displaying the IP address of a remote NFS/RDMA
client.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Commit d21b05f1 ("rdma: SVCRMDA Header File")
introduced the SVCRDMA_DEBUG macro, but it doesn't seem to have been
used.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Use a consistent naming convention so that these trace
points can be enabled quickly via a glob.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Way back when I was writing the RPC/RDMA server-side backchannel
code, I misread the TCP backchannel reply handler logic. When
svc_tcp_recvfrom() successfully receives a backchannel reply, it
does not return -EAGAIN. It sets XPT_DATA and returns zero.

Update svc_rdma_recvfrom() to return zero. Here, XPT_DATA doesn't
need to be set again: it is set whenever a new message is received,
behind a spin lock in a single threaded context.

Also, if handling the cb reply is not successful, the message is
simply dropped. There's no special message framing to deal with as
there is in the TCP case.

Now that the handle_bc_reply() return value is ignored, I've removed
the dprintk call sites in the error exit of handle_bc_reply() in
favor of trace points in other areas that already report the error
cases.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Replace a dprintk call site.

This is the last remaining dprintk call site in svc_rdma_rw.c, so
remove dprintk infrastructure as well.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Replace a dprintk call site with a tracepoint.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Clean up: Replace two dprintk call sites with a tracepoint.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

- De-duplicate code
- Rename the tracepoint with "_err" to allow enabling via glob
- Report the sg_cnt for the failing rw_ctx
- Fix a dumb signage issue
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

It appears that the RPC/RDMA transport does not need serialization
of calls to its xpo_sendto method. Move the mutex into the socket
methods that still need that serialization.

Tail latencies are unambiguously better with this patch applied.
fio randrw 8KB 70/30 on NFSv3, smaller numbers are better:

    clat percentiles (usec):

With xpt_mutex:
r    | 99.99th=[ 8848]
w    | 99.99th=[ 9634]

Without xpt_mutex:
r    | 99.99th=[ 8586]
w    | 99.99th=[ 8979]

Serializing the construction of RPC/RDMA transport headers is not
really necessary at this point, because the Linux NFS server
implementation never changes its credit grant on a connection. If
that should change, then svc_rdma_sendto will need to serialize
access to the transport's credit grant fields.
Reported-by: kbuild test robot <lkp@intel.com>
[ cel: fix uninitialized variable warning ]
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Pull IPMI update from Corey Minyard:
 "Convert i2c_new_device() to i2c_new_client_device()

  Wolfram Sang has asked to have this included in 5.7 so the deprecated
  API can be removed next release. There should be no functional
  difference.

  I think that entire this section of code can be removed; it is
  leftover from other things that have since changed, but this is the
  safer thing to do for now. The full removal can happen next release"

* tag 'for-linus-5.7-2' of git://github.com/cminyard/linux-ipmi:
  char: ipmi: convert to use i2c_new_client_device()

Pull clk fixes from Stephen Boyd:
 "Some more clk driver fixes and one core framework fix:

   - A handful of TI driver fixes for bad of_node_put() and incorrect
     parent names

   - Rockchip rk3228 aclk_gpu* creation was interfering with lima GPU
     work so we use a composite clk now

   - Resuming from suspend on Tegra Jetson TK1 was broken because an
     audio PLL calculated an incorrect rate

   - A fix for devicetree probing on IM-PD1 by actually specifying a clk
     name which is required to pass clk registration

   - Avoid list corruption if registration fails for a critical clk"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: ti: clkctrl: convert subclocks to use proper names also
  clk: ti: am33xx: fix RTC clock parent
  clk: ti: clkctrl: Fix Bad of_node_put within clkctrl_get_name
  clk: tegra: Fix initial rate for pll_a on Tegra124
  clk: impd1: Look up clock-output-names
  clk: Unlink clock if failed to prepare or enable
  clk: rockchip: fix incorrect configuration of rk3228 aclk_gpu* clocks

Pull USB fixes from Greg KH:
 "Here are a number of USB fixes for 5.7-rc6

  The "largest" in here is a bunch of raw-gadget fixes and api changes
  as the driver just showed up in -rc1 and work has been done to fix up
  some uapi issues found with the original submission, before it shows
  up in a -final release.

  Other than that, a bunch of other small USB gadget fixes, xhci fixes,
  some quirks, andother tiny fixes for reported issues.

  All of these have been in linux-next with no reported issues"

* tag 'usb-5.7-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (26 commits)
  USB: gadget: fix illegal array access in binding with UDC
  usb: core: hub: limit HUB_QUIRK_DISABLE_AUTOSUSPEND to USB5534B
  USB: usbfs: fix mmap dma mismatch
  usb: host: xhci-plat: keep runtime active when removing host
  usb: xhci: Fix NULL pointer dereference when enqueuing trbs from urb sg list
  usb: cdns3: gadget: make a bunch of functions static
  usb: mtu3: constify struct debugfs_reg32
  usb: gadget: udc: atmel: Make some symbols static
  usb: raw-gadget: fix null-ptr-deref when reenabling endpoints
  usb: raw-gadget: documentation updates
  usb: raw-gadget: support stalling/halting/wedging endpoints
  usb: raw-gadget: fix gadget endpoint selection
  usb: raw-gadget: improve uapi headers comments
  usb: typec: mux: intel: Fix DP_HPD_LVL bit field
  usb: raw-gadget: fix return value of ep read ioctls
  usb: dwc3: select USB_ROLE_SWITCH
  usb: gadget: legacy: fix error return code in gncm_bind()
  usb: gadget: legacy: fix error return code in cdc_bind()
  usb: gadget: legacy: fix redundant initialization warnings
  usb: gadget: tegra-xudc: Fix idle suspend/resume
  ...

Pull execve fix from Eric Biederman:
 "While working on my exec cleanups I found a bug in exec that I
  introduced by accident a couple of years ago. I apparently missed the
  fact that bprm->file can change.

  Now I have a very personal motive to clean up exec and make it more
  approachable.

  The change is just moving woud_dump to where it acts on the final
  bprm->file not the initial bprm->file. I have been careful and tested
  and verify this fix works"

* 'exec-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
  exec: Move would_dump into flush_old_exec

Pull x86 stack unwinding fix from Thomas Gleixner:
 "A single bugfix for the ORC unwinder to ensure that the error flag
  which tells the unwinding code whether a stack trace can be trusted or
  not is always set correctly.

  This was messed up by a couple of changes in the recent past"

* tag 'objtool-urgent-2020-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/unwind/orc: Fix error handling in __unwind_start()

Pull x86 fix from Borislav Petkov:
 "A single fix for early boot crashes of kernels built with gcc10 and
  stack protector enabled"

* tag 'x86_urgent_for_v5.7-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86: Fix early boot crash on gcc-10, third try

I goofed when I added mm->user_ns support to would_dump.  I missed the
fact that in the case of binfmt_loader, binfmt_em86, binfmt_misc, and
binfmt_script bprm->file is reassigned.  Which made the move of
would_dump from setup_new_exec to __do_execve_file before exec_binprm
incorrect as it can result in would_dump running on the script instead
of the interpreter of the script.

The net result is that the code stopped making unreadable interpreters
undumpable.  Which allows them to be ptraced and written to disk
without special permissions.  Oops.

The move was necessary because the call in set_new_exec was after
bprm->mm was no longer valid.

To correct this mistake move the misplaced would_dump from
__do_execve_file into flos_old_exec, before exec_mmap is called.

I tested and confirmed that without this fix I can attach with gdb to
a script with an unreadable interpreter, and with this fix I can not.

Cc: stable@vger.kernel.org
Fixes: f84df2a6 ("exec: Ensure mm->user_ns contains the execed files")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Pull cifs fixes from Steve French:
 "Three small cifs/smb3 fixes, one for stable"

* tag '5.7-rc5-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix leaked reference on requeued write
  cifs: Fix null pointer check in cifs_read
  CIFS: Spelling s/EACCESS/EACCES/

Pull kvm fixes from Paolo Bonzini:
 "A new testcase for guest debugging (gdbstub) that exposed a bunch of
  bugs, mostly for AMD processors. And a few other x86 fixes"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: Fix off-by-one error in kvm_vcpu_ioctl_x86_setup_mce
  KVM: x86: Fix pkru save/restore when guest CR4.PKE=0, move it to x86.c
  KVM: SVM: Disable AVIC before setting V_IRQ
  KVM: Introduce kvm_make_all_cpus_request_except()
  KVM: VMX: pass correct DR6 for GD userspace exit
  KVM: x86, SVM: isolate vcpu->arch.dr6 from vmcb->save.dr6
  KVM: SVM: keep DR6 synchronized with vcpu->arch.dr6
  KVM: nSVM: trap #DB and #BP to userspace if guest debugging is on
  KVM: selftests: Add KVM_SET_GUEST_DEBUG test
  KVM: X86: Fix single-step with KVM_SET_GUEST_DEBUG
  KVM: X86: Set RTM for DB_VECTOR too for KVM_EXIT_DEBUG
  KVM: x86: fix DR6 delivery for various cases of #DB injection
  KVM: X86: Declare KVM_CAP_SET_GUEST_DEBUG properly