[PATCH 2/3] lib: sbi_ipi: Process self-IPIs in sbi_ipi_send()

[Samuel Holland]

An IPI sent to the local hart can be processed directly instead of
triggering the IPI device. This is more efficient, and it avoids a
deadlock when the .sync callback is defined. Since interrupts are
disabled while handling an ecall, the IPI would not get delivered
until the next mret, but sbi_ipi_sync() is called before then.

Signed-off-by: Samuel Holland <samuel.holland at sifive.com>
---

 lib/sbi/sbi_ipi.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/lib/sbi/sbi_ipi.c b/lib/sbi/sbi_ipi.c
index 1fd24228..ac31cd94 100644
--- a/lib/sbi/sbi_ipi.c
+++ b/lib/sbi/sbi_ipi.c
@@ -60,6 +60,14 @@ static int sbi_ipi_send(struct sbi_scratch *scratch, u32 remote_hartindex,
 				      remote_hartindex, data);
 		if (ret != SBI_IPI_UPDATE_SUCCESS)
 			return ret;
+	} else if (scratch == remote_scratch) {
+		/*
+		 * IPI events with an update() callback are expected to return
+		 * SBI_IPI_UPDATE_BREAK for self-IPIs. For other events, check
+		 * for self-IPI and execute the callback directly here.
+		 */
+		ipi_ops->process(scratch);
+		return 0;
 	}
 
 	/*
-- 
2.42.0