11.04.2019 17:47, Eric Blake wrote: > On 4/11/19 2:57 AM, Vladimir Sementsov-Ogievskiy wrote: >> 10.04.2019 23:42, Eric Blake wrote: >>> Filter the qemu-nbd server output to get rid of a direct reference >>> to my build directory. >>> >>> Fixes: e9dce9cb >>> Reported-by: Max Reitz <mre...@redhat.com> >>> Signed-off-by: Eric Blake <ebl...@redhat.com> >>> --- >>> >>> Not worth -rc4 on its own, but if something else pops up that requires >>> another spin, I plan on a pull request for this one. Otherwise it >>> slips to 4.1, and 4.0 just has a broken iotest. >>> > >>> @@ -69,12 +70,13 @@ echo >>> >>> # Intentionally omit '-f' to force image probing, which in turn forces >>> # sector alignment, here at the server. >>> -nbd_server_start_unix_socket "$TEST_IMG_FILE" >>> +nbd_server_start_unix_socket "$TEST_IMG_FILE" 2> "$TEST_DIR/server.log" >>> > >>> 1 KiB (0x400) bytes allocated at offset 0 bytes (0x0) >>> +WARNING: Image format was not specified for 'TEST_DIR/t.raw' and probing >>> guessed raw. >>> + Automatically detecting the format is dangerous for raw images, >>> write operations on block 0 will be restricted. >>> + Specify the 'raw' format explicitly to remove the restrictions. >> >> May be, just add -f raw to nbd server, to make this message disappear? And >> than, don't do any redirections, as if there >> will be failed unfiltered output it is OK? > > No. See the comment in the test - it is intentional that we are omitting > -f, in two different places, in order to get three test scenarios: > > normal: server '-f raw' 1-byte align, client '-f raw' 1-byte align > server-constrained: server omit for 512-byte align, client '-f raw' > obeys server align > client-constrained: server '-f raw' 1-byte align, client omit for > intentional 512-byte align > > in order to show both that the server with forced alignment is able to > round requests correctly before sending to the client, and to show that > the client with forced alignment is able to round requests correctly in > spite of the server sending answers narrower than the client wants. > > In 4.1, when Max's filter handling patches land, it will be possible to > use the blkdebug driver to force alignment instead of omitting '-f raw'; > at which point we can take away the server.log file. But until the > filter handling patches land, any use of blkdebug causes odd test > misbehaviors because not everything in the block layer properly sees > through filter nodes. >
Understand now, thanks. Reviewed-by: Vladimir Sementsov-Ogievskiy <vsement...@virtuozzo.com> -- Best regards, Vladimir
