| Message ID | 87a9x1ls3b.fsf@elfo.mitica |
|---|---|
| State | New |
| Headers | show |
Il 07/09/2012 18:23, Juan Quintela ha scritto: > > Hi > > here is v2 of the migration thread series. There is still some "issues" > with locking in the error paths (they are at 54 patches now). > > Changes from v1: > - migration stats series are included > - migration bitmap sync trace-events to know how long it takes > - file->last_error use almost removed > reworked functions to return real error codes and work with that. > Some more work needed here. > - new savevm for live migration pending method. see last commit for > details. Can you start factoring out any cleanup that can be applied independently? Paolo > Please test and comment. > > Later, Juan. > > The following changes since commit 6e4c0d1f03d6ab407509c32fab7cb4b8230f57ff: > > hw/pl110: Fix spelling of 'palette' (2012-09-06 17:04:33 +0200) > > are available in the git repository at: > > http://repo.or.cz/r/qemu/quintela.git migration-thread-v2 > > for you to fetch changes up to 688feac0fbc287920dff537ed13fb8483c064f7f: > > savem: Add calculating a new save_live migration method: pending (2012-09-07 14:00:35 +0200) > > ---------------------------------------------------------------- > Juan Quintela (49): > buffered_file: g_realloc() can't fail > fix migration sync > migration: store end_time in a local variable > migration: print total downtime for final phase of migration > migration: rename expected_time to expected_downtime > migration: export migrate_get_current() > migration: print expected downtime in info migrate > savevm: Factorize ram globals reset in its own function > ram: introduce migration_bitmap_set_dirty() > ram: Introduce migration_bitmap_test_and_reset_dirty() > ram: Export last_ram_offset() > ram: introduce migration_bitmap_sync() > ram: create trace event for migration sync bitmap > Separate migration bitmap > migration: Add dirty_pages_rate to query migrate output > buffered_file: rename opaque to migration_state > buffered_file: opaque is MigrationState > buffered_file: unfold migrate_fd_put_buffer > buffered_file: unfold migrate_fd_put_ready > buffered_file: unfold migrate_fd_put_buffer > buffered_file: unfold migrate_fd_put_buffer > buffered_file: We can access directly to bandwidth_limit > buffered_file: callers of buffered_flush() already check for errors > buffered_file: make buffered_flush return the error code > migration: make migrate_fd_wait_for_unfreeze() return errors > savevm: unexport qemu_fflush > viritio-net: use qemu_get_buffer() in a temp buffer > savevm: Remove qemu_fseek() > savevm: make qemu_fflush() return an error code > savevm: unfold qemu_fclose_internal() > savevm: unexport qemu_ftell() > savevm: make qemu_fill_buffer() be consistent > savevm: Only qemu_fflush() can generate errors > buffered_file: buffered_put_buffer() don't need to set last_error > block-migration: make flush_blks() return errors > block-migration: Switch meaning of return value > block-migration: handle errors with the return codes correctly > savevm: un-export qemu_file_set_error() > savevm: make qemu_file_put_notify() return errors > buffered_file: Move from using a timer to use a thread > migration: make qemu_fopen_ops_buffered() return void > migration: stop all cpus correctly > migration: make writes blocking > migration: remove unfreeze logic > migration: take finer locking > buffered_file: Unfold the trick to restart generating migration data > buffered_file: don't flush on put buffer > buffered_file: unfold buffered_append in buffered_put_buffer > savem: Add calculating a new save_live migration method: pending > > Paolo Bonzini (2): > split MRU ram list > BufferedFile: append, then flush > > Umesh Deshpande (2): > add a version number to ram_list > protect the ramlist with a separate mutex > > arch_init.c | 174 ++++++++++++++++++++++++++++++++------------ > block-migration.c | 100 +++++++++---------------- > buffered_file.c | 213 +++++++++++++++++++++--------------------------------- > buffered_file.h | 12 +-- > cpu-all.h | 17 ++++- > exec-obsolete.h | 10 --- > exec.c | 45 ++++++++++-- > hmp.c | 12 +++ > hw/virtio-net.c | 4 +- > migration-exec.c | 2 - > migration-fd.c | 6 -- > migration-tcp.c | 2 +- > migration-unix.c | 2 - > migration.c | 151 +++++++++++++++++++------------------- > migration.h | 10 +++ > qapi-schema.json | 18 ++++- > qemu-file.h | 11 --- > qmp-commands.hx | 9 +++ > savevm.c | 144 ++++++++++++++++++------------------ > sysemu.h | 1 + > trace-events | 4 + > vmstate.h | 1 + > 22 files changed, 498 insertions(+), 450 deletions(-) > >
Paolo Bonzini <pbonzini@redhat.com> wrote: > Il 07/09/2012 18:23, Juan Quintela ha scritto: >> >> Hi >> >> here is v2 of the migration thread series. There is still some "issues" >> with locking in the error paths (they are at 54 patches now). >> >> Changes from v1: >> - migration stats series are included >> - migration bitmap sync trace-events to know how long it takes >> - file->last_error use almost removed >> reworked functions to return real error codes and work with that. >> Some more work needed here. >> - new savevm for live migration pending method. see last commit for >> details. > > Can you start factoring out any cleanup that can be applied independently? This is the breakpoint. buffered_file: Move from using a timer to use a thread Everything before this could be integrated Today, from there, it needs still more 'thought". As said, buffered_file.c is almost gone, and file->last_error is almost gone. That were the things that I wanted to clean to be able to split the generation of migration data and its writting. Later, Juan.
Il 07/09/2012 21:52, Juan Quintela ha scritto: > This is the breakpoint. > > buffered_file: Move from using a timer to use a thread > > Everything before this could be integrated Today, from there, it needs > still more 'thought". True, on the other hand, the split ramlist mutex etc. would be premature. Paolo
Hi here is v2 of the migration thread series. There is still some "issues" with locking in the error paths (they are at 54 patches now). Changes from v1: - migration stats series are included - migration bitmap sync trace-events to know how long it takes - file->last_error use almost removed reworked functions to return real error codes and work with that. Some more work needed here. - new savevm for live migration pending method. see last commit for details. Please test and comment. Later, Juan. The following changes since commit 6e4c0d1f03d6ab407509c32fab7cb4b8230f57ff: hw/pl110: Fix spelling of 'palette' (2012-09-06 17:04:33 +0200) are available in the git repository at: http://repo.or.cz/r/qemu/quintela.git migration-thread-v2 for you to fetch changes up to 688feac0fbc287920dff537ed13fb8483c064f7f: savem: Add calculating a new save_live migration method: pending (2012-09-07 14:00:35 +0200) ---------------------------------------------------------------- Juan Quintela (49): buffered_file: g_realloc() can't fail fix migration sync migration: store end_time in a local variable migration: print total downtime for final phase of migration migration: rename expected_time to expected_downtime migration: export migrate_get_current() migration: print expected downtime in info migrate savevm: Factorize ram globals reset in its own function ram: introduce migration_bitmap_set_dirty() ram: Introduce migration_bitmap_test_and_reset_dirty() ram: Export last_ram_offset() ram: introduce migration_bitmap_sync() ram: create trace event for migration sync bitmap Separate migration bitmap migration: Add dirty_pages_rate to query migrate output buffered_file: rename opaque to migration_state buffered_file: opaque is MigrationState buffered_file: unfold migrate_fd_put_buffer buffered_file: unfold migrate_fd_put_ready buffered_file: unfold migrate_fd_put_buffer buffered_file: unfold migrate_fd_put_buffer buffered_file: We can access directly to bandwidth_limit buffered_file: callers of buffered_flush() already check for errors buffered_file: make buffered_flush return the error code migration: make migrate_fd_wait_for_unfreeze() return errors savevm: unexport qemu_fflush viritio-net: use qemu_get_buffer() in a temp buffer savevm: Remove qemu_fseek() savevm: make qemu_fflush() return an error code savevm: unfold qemu_fclose_internal() savevm: unexport qemu_ftell() savevm: make qemu_fill_buffer() be consistent savevm: Only qemu_fflush() can generate errors buffered_file: buffered_put_buffer() don't need to set last_error block-migration: make flush_blks() return errors block-migration: Switch meaning of return value block-migration: handle errors with the return codes correctly savevm: un-export qemu_file_set_error() savevm: make qemu_file_put_notify() return errors buffered_file: Move from using a timer to use a thread migration: make qemu_fopen_ops_buffered() return void migration: stop all cpus correctly migration: make writes blocking migration: remove unfreeze logic migration: take finer locking buffered_file: Unfold the trick to restart generating migration data buffered_file: don't flush on put buffer buffered_file: unfold buffered_append in buffered_put_buffer savem: Add calculating a new save_live migration method: pending Paolo Bonzini (2): split MRU ram list BufferedFile: append, then flush Umesh Deshpande (2): add a version number to ram_list protect the ramlist with a separate mutex arch_init.c | 174 ++++++++++++++++++++++++++++++++------------ block-migration.c | 100 +++++++++---------------- buffered_file.c | 213 +++++++++++++++++++++--------------------------------- buffered_file.h | 12 +-- cpu-all.h | 17 ++++- exec-obsolete.h | 10 --- exec.c | 45 ++++++++++-- hmp.c | 12 +++ hw/virtio-net.c | 4 +- migration-exec.c | 2 - migration-fd.c | 6 -- migration-tcp.c | 2 +- migration-unix.c | 2 - migration.c | 151 +++++++++++++++++++------------------- migration.h | 10 +++ qapi-schema.json | 18 ++++- qemu-file.h | 11 --- qmp-commands.hx | 9 +++ savevm.c | 144 ++++++++++++++++++------------------ sysemu.h | 1 + trace-events | 4 + vmstate.h | 1 + 22 files changed, 498 insertions(+), 450 deletions(-)