Explicitly print errors about -EAGAIN in io_uring

Vitaliy Filippov 2022-12-17 15:49:49 +03:00
parent 5464821fa5
commit 552e207d2b
5 changed files with 18 additions and 20 deletions

View File

@ -35,24 +35,14 @@ journal_flusher_co::journal_flusher_co()
{
bs->live = true;
if (data->res != data->iov.iov_len)
{
throw std::runtime_error(
"data read operation failed during flush ("+std::to_string(data->res)+" != "+std::to_string(data->iov.iov_len)+
"). can't continue, sorry :-("
);
}
bs->disk_error_abort("read operation during flush", data->res, data->iov.iov_len);
wait_count--;
};
simple_callback_w = [this](ring_data_t* data)
{
bs->live = true;
if (data->res != data->iov.iov_len)
{
throw std::runtime_error(
"write operation failed ("+std::to_string(data->res)+" != "+std::to_string(data->iov.iov_len)+
"). state "+std::to_string(wait_state)+". in-memory state is corrupted. AAAAAAAaaaaaaaaa!!!111"
);
}
bs->disk_error_abort("write operation during flush", data->res, data->iov.iov_len);
wait_count--;
};
}

View File

@ -676,3 +676,16 @@ void blockstore_impl_t::dump_diagnostics()
journal.dump_diagnostics();
flusher->dump_diagnostics();
}
void blockstore_impl_t::disk_error_abort(const char *op, int retval, int expected)
{
if (retval == -EAGAIN)
{
fprintf(stderr, "EAGAIN error received from a disk %s during flush."
" It must never happen with io_uring and indicates a kernel bug."
" Please upgrade your kernel. Aborting.\n", op);
exit(1);
}
fprintf(stderr, "Disk %s failed: result is %d, expected %d. Can't continue, sorry :-(\n", op, retval, expected);
exit(1);
}

View File

@ -292,6 +292,7 @@ class blockstore_impl_t
// Journaling
void prepare_journal_sector_write(int sector, blockstore_op_t *op);
void handle_journal_write(ring_data_t *data, uint64_t flush_id);
void disk_error_abort(const char *op, int retval, int expected);
// Asynchronous init
int initialized;

View File

@ -198,10 +198,7 @@ void blockstore_impl_t::handle_journal_write(ring_data_t *data, uint64_t flush_i
if (data->res != data->iov.iov_len)
{
// FIXME: our state becomes corrupted after a write error. maybe do something better than just die
throw std::runtime_error(
"journal write failed ("+std::to_string(data->res)+" != "+std::to_string(data->iov.iov_len)+
"). in-memory state is corrupted. AAAAAAAaaaaaaaaa!!!111"
);
disk_error_abort("journal write", data->res, data->iov.iov_len);
}
auto fl_it = journal.flushing_ops.upper_bound((pending_journaling_t){ .flush_id = flush_id });
if (fl_it != journal.flushing_ops.end() && fl_it->flush_id == flush_id)

View File

@ -586,10 +586,7 @@ void blockstore_impl_t::handle_write_event(ring_data_t *data, blockstore_op_t *o
if (data->res != data->iov.iov_len)
{
// FIXME: our state becomes corrupted after a write error. maybe do something better than just die
throw std::runtime_error(
"write operation failed ("+std::to_string(data->res)+" != "+std::to_string(data->iov.iov_len)+
"). in-memory state is corrupted. AAAAAAAaaaaaaaaa!!!111"
);
disk_error_abort("data write", data->res, data->iov.iov_len);
}
PRIV(op)->pending_ops--;
assert(PRIV(op)->pending_ops >= 0);