Skip to contentSkip to conversation

hw/char/exynos4210_uart: Fix crash on trying to load VM state

PATCH In review Peter Maydell 3 participants 4 messages Jan 20, 2022 – Jan 22, 2022 Arm Samsung Exynos

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hw/char/exynos4210_uart: Fix crash on trying to load VM state

The exynos4210_uart_post_load() function assumes that it is passed the Exynos4210UartState, but it has been attached to the VMStateDescription for the Exynos4210UartFIFO type. The result is a SIGSEGV when attempting to load VM state for any machine type including this device.

Fix the bug by attaching the post-load function to the VMSD for the Exynos4210UartState. This is the logical place for it, because the actions it does relate to the entire UART state, not just the FIFO.

Thanks to the bug reporter @TrungNguyen1909 for the clear bug description and the suggested fix.

Fixes c9d3396d80fe7ece9b

("hw/char/exynos4210_uart: Implement post_load function")

---
 hw/char/exynos4210_uart.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/char/exynos4210_uart.c +1 -1
index 80d401a3795..addcd59b028 100644
@@ -628,7 +628,6 @@ static const VMStateDescription vmstate_exynos4210_uart_fifo = {
     .name = "exynos4210.uart.fifo",
     .version_id = 1,
     .minimum_version_id = 1,
-    .post_load = exynos4210_uart_post_load,
     .fields = (VMStateField[]) {
         VMSTATE_UINT32(sp, Exynos4210UartFIFO),
         VMSTATE_UINT32(rp, Exynos4210UartFIFO),
@@ -641,6 +640,7 @@ static const VMStateDescription vmstate_exynos4210_uart = {
     .name = "exynos4210.uart",
     .version_id = 1,
     .minimum_version_id = 1,
+    .post_load = exynos4210_uart_post_load,
     .fields = (VMStateField[]) {
         VMSTATE_STRUCT(rx, Exynos4210UartState, 1,
                        vmstate_exynos4210_uart_fifo, Exynos4210UartFIFO),
Signature · 1 line
2.25.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 10 lines · Peter Maydell

The exynos4210_uart_post_load() function assumes that it is passed the Exynos4210UartState, but it has been attached to the VMStateDescription for the Exynos4210UartFIFO type. The result is a SIGSEGV when attempting to load VM state for any machine type including this device.

Fix the bug by attaching the post-load function to the VMSD for the Exynos4210UartState. This is the logical place for it, because the actions it does relate to the entire UART state, not just the FIFO.

Quoted 7 lines

Thanks to the bug reporter @TrungNguyen1909 for the clear bug description and the suggested fix.

Fixes c9d3396d80fe7ece9b

("hw/char/exynos4210_uart: Implement post_load function")

Quoted 25 lines
---
  hw/char/exynos4210_uart.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
hw/char/exynos4210_uart.c +1 -1
index 80d401a3795..addcd59b028 100644
@@ -628,7 +628,6 @@ static const VMStateDescription vmstate_exynos4210_uart_fifo = {
      .name = "exynos4210.uart.fifo",
      .version_id = 1,
      .minimum_version_id = 1,
-    .post_load = exynos4210_uart_post_load,
      .fields = (VMStateField[]) {
          VMSTATE_UINT32(sp, Exynos4210UartFIFO),
          VMSTATE_UINT32(rp, Exynos4210UartFIFO),
@@ -641,6 +640,7 @@ static const VMStateDescription vmstate_exynos4210_uart = {
      .name = "exynos4210.uart",
      .version_id = 1,
      .minimum_version_id = 1,
+    .post_load = exynos4210_uart_post_load,
      .fields = (VMStateField[]) {
          VMSTATE_STRUCT(rx, Exynos4210UartState, 1,
                         vmstate_exynos4210_uart_fifo, Exynos4210UartFIFO),
Reviewed-by Philippe Mathieu-Daudé on v1 1/1
#2 lore raw
In reply to Peter Maydell
Quoted 10 lines · Peter Maydell

The exynos4210_uart_post_load() function assumes that it is passed the Exynos4210UartState, but it has been attached to the VMStateDescription for the Exynos4210UartFIFO type. The result is a SIGSEGV when attempting to load VM state for any machine type including this device.

Fix the bug by attaching the post-load function to the VMSD for the Exynos4210UartState. This is the logical place for it, because the actions it does relate to the entire UART state, not just the FIFO.

Quoted 6 lines

Thanks to the bug reporter @TrungNguyen1909 for the clear bug description and the suggested fix.

Fixes c9d3396d80fe7ece9b

("hw/char/exynos4210_uart: Implement post_load function")

Apparently GitLab doesn't recognize "Buglink":
https://docs.gitlab.com/ee/user/project/issues/managing_issues.html#default-closing-pattern
which might be why we use "Resolves: " to have GitLab automatically close issues.

Otherwise:

Quoted 4 lines
---
  hw/char/exynos4210_uart.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
#3 lore raw
Quoted 15 lines · Philippe Mathieu-Daudé
Quoted 13 lines · Peter Maydell

The exynos4210_uart_post_load() function assumes that it is passed the Exynos4210UartState, but it has been attached to the VMStateDescription for the Exynos4210UartFIFO type. The result is a SIGSEGV when attempting to load VM state for any machine type including this device.

Fix the bug by attaching the post-load function to the VMSD for the Exynos4210UartState. This is the logical place for it, because the actions it does relate to the entire UART state, not just the FIFO.

Thanks to the bug reporter @TrungNguyen1909 for the clear bug description and the suggested fix.

Quoted 8 lines
Quoted 3 lines
Fixes c9d3396d80fe7ece9b

("hw/char/exynos4210_uart: Implement post_load function")

Apparently GitLab doesn't recognize "Buglink":
https://docs.gitlab.com/ee/user/project/issues/managing_issues.html#default-closing-pattern
which might be why we use "Resolves: " to have GitLab automatically close issues.

Thanks; I can never remember which tag is the right one. I think I just fished this one out of a random commit in the git history, but I should have checked the docs, where we do have this documented.

-- PMM