On Feb 16, 2023 / 19:10, Ming Lei wrote:
> On Thu, Feb 16, 2023 at 04:58:00PM +0800, Ming Lei wrote:
> > Hi Shinichiro,
> >
> > Thanks for the review!
> >
> > On Thu, Feb 16, 2023 at 08:19:39AM +0000, Shinichiro Kawasaki wrote:
> > > Hi Ming, thanks for the patches. It sounds a good idea to extend blktests
> > > coverage to ublk.
> > >
> > > Regarding the commit title, I suggest this:
> > >
> > > src: add mini ublk source codes
> > >
> > > The word "blktests" can be placed after the word "PATCH" as follows:
> > >
> > > [PATCH blktests] src: add mini ublk source codes
> > >
> > > Please try --subject-prefix="PATCH blktests" option for git format-patch.
> >
> > OK.
> >
> > >
> > > On Feb 16, 2023 / 11:01, Ming Lei wrote:
> > > > Prepare for adding ublk related test:
> > > >
> > > > 1) ublk delete is sync removal, this way is convenient to
> > > > blkg/queue/disk instance leak issue
> > > >
> > > > 2) mini ublk has two builtin target(null, loop), and loop IO is
> > > > handled by io_uring, so we can use ublk to cover part of io_uring
> > > > workloads
> > > >
> > > > 3) not like loop/nbd, ublk won't pre-allocate/add disk, and always
> > > > add/delete disk dynamically, this way may cover disk plug & unplug
> > > > tests
> > > >
> > > > 4) ublk specific test given people starts to use it, so better to
> > > > let blktest cover ublk related tests
> > > >
> > > > Add mini ublk source for test purpose only, which is easy to use:
> > > >
> > > > ./miniublk add -t {null|loop} [-q nr_queues] [-d depth] [-n dev_id]
> > > > default: nr_queues=2(max 4), depth=128(max 128), dev_id=-1(auto allocation)
> > > > -t loop -f backing_file
> > > > -t null
> > > > ./miniublk del [-n dev_id] [--disk/-d disk_path ] -a
> > > > -a delete all devices, -n delete specified device
> > > > ./miniublk list [-n dev_id] -a
> > > > -a list all devices, -n list specified device, default -a
> > > >
> > > > ublk depends on liburing 2.2, so allow to ignore ublk build failure
> > > > in case of missing liburing, and we will check if ublk program exits
> > > > inside test.
> > > >
> > > > Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx>
> > > > ---
> > > > Makefile | 2 +-
> > > > src/Makefile | 13 +-
> > > > src/ublk/miniublk.c | 1385 +++++++++++++++++++++++++++++++++++++++++++
> > > > src/ublk/ublk_cmd.h | 278 +++++++++
> > > > 4 files changed, 1674 insertions(+), 4 deletions(-)
> > > > create mode 100644 src/ublk/miniublk.c
> > > > create mode 100644 src/ublk/ublk_cmd.h
> > > >
> > > > diff --git a/Makefile b/Makefile
> > > > index 5a04479..b9bbade 100644
> > > > --- a/Makefile
> > > > +++ b/Makefile
> > > > @@ -2,7 +2,7 @@ prefix ?= /usr/local
> > > > dest = $(DESTDIR)$(prefix)/blktests
> > > >
> > > > all:
> > > > - $(MAKE) -C src all
> > > > + $(MAKE) -i -C src all
> > >
> > > This -i option to ignore errors is applied to all build targets, so it will hide
> > > errors. Instead os ignoring the error, how about checking the liburing version
> > > with pkg-config command? I roughly implemented it as the patch below on top of
> > > your patch. It looks working.
> > >
> > > diff --git a/src/Makefile b/src/Makefile
> > > index 9bb8da6..c600099 100644
> > > --- a/src/Makefile
> > > +++ b/src/Makefile
> > > @@ -2,6 +2,11 @@ HAVE_C_HEADER = $(shell if echo "\#include <$(1)>" | \
> > > $(CC) -E - > /dev/null 2>&1; then echo "$(2)"; \
> > > else echo "$(3)"; fi)
> > >
> > > +HAVE_PACKAGE_VER = $(shell if pkg-config --atleast-version="$(2)" "$(1)"; \
> > > + then echo 1; else echo 0; fi)
> > > +
> > > +HAVE_LIBURING := $(call HAVE_PACKAGE_VER,liburing,2.2)
> >
> > I tried this way, and it fails in case that liburing is built
> > against source tree directly. And liburing2.2 is still a bit new, and even
> > some distributions doesn't package it. I will think about other way
> > for the check.
>
> Looks the following way works:
>
>
> diff --git a/src/Makefile b/src/Makefile
> index 83e8a62..adfd27a 100644
> --- a/src/Makefile
> +++ b/src/Makefile
> @@ -2,6 +2,10 @@ HAVE_C_HEADER = $(shell if echo "\#include <$(1)>" | \
> $(CC) -E - > /dev/null 2>&1; then echo "$(2)"; \
> else echo "$(3)"; fi)
>
> +HAVE_C_MACRO = $(shell if echo "#include <$(1)>" | \
> + $(CC) -E - | grep $(2) > /dev/null 2>&1; then echo 1; \
> + else echo 0; fi)
Oh, this macro check idea looks nifty :) It looks working in my environment too.
When /usr/include/liburing.h does not exist, $(CC) spits out an error. It would
be the better add one more stderr redirect to /dev/null.
HAVE_C_MACRO = $(shell if echo "#include <$(1)>" | $(CC) -E - 2>&1 /dev/null \
| grep $(2) > /dev/null 2>&1; then echo 1; else echo 0; fi)
> +
> C_TARGETS := \
> loblksize \
> loop_change_fd \
> @@ -15,25 +19,30 @@ C_TARGETS := \
>
> C_MINIUBLK := ublk/miniublk
>
> +HAVE_LIBURING := $(call HAVE_C_MACRO,liburing.h,IORING_OP_URING_CMD)
> +
> CXX_TARGETS := \
> discontiguous-io
>
> +ifeq ($(HAVE_LIBURING), 1)
> +TARGETS := $(C_TARGETS) $(CXX_TARGETS) $(C_MINIUBLK)
> +else
> TARGETS := $(C_TARGETS) $(CXX_TARGETS)
> -ALL_TARGETS := $(TARGETS) $(C_MINIUBLK)
I suggest to add a message here, so that users can tell that miniublk was not
built.
$(info Skip $(C_MINIUBLK) build due to old or no liburing)
> +endif
>
> CONFIG_DEFS := $(call HAVE_C_HEADER,linux/blkzoned.h,-DHAVE_LINUX_BLKZONED_H)
>
> override CFLAGS := -O2 -Wall -Wshadow $(CFLAGS) $(CONFIG_DEFS)
> override CXXFLAGS := -O2 -std=c++11 -Wall -Wextra -Wshadow -Wno-sign-compare \
> -Werror $(CXXFLAGS) $(CONFIG_DEFS)
> -MINIUBLK_FLAGS := -D_GNU_SOURCE -lpthread -luring
> +MINIUBLK_FLAGS := -D_GNU_SOURCE -lpthread -luring -Iublk
>
> -all: $(ALL_TARGETS)
> +all: $(TARGETS)
>
> clean:
> - rm -f $(ALL_TARGETS)
> + rm -f $(TARGETS)
>
> -install: $(ALL_TARGETS)
> +install: $(TARGETS)
> install -m755 -d $(dest)
> install $(TARGETS) $(dest)
>
>
>
> Thanks,
> Ming
>
--
Shin'ichiro Kawasaki
