Dear Linux folks,
Could you please apply commit 0c25422d34b4 (scsi: mpt3sas: Remove scsi_dma_map() error messages) to the 5.15.y series?
commit 0c25422d34b4726b2707d5f38560943155a91b80 Author: Sreekanth Reddy sreekanth.reddy@broadcom.com Date: Thu Mar 3 19:32:03 2022 +0530
scsi: mpt3sas: Remove scsi_dma_map() error messages
When scsi_dma_map() fails by returning a sges_left value less than zero, the amount of logging produced can be extremely high. In a recent end-user environment, 1200 messages per second were being sent to the log buffer. This eventually overwhelmed the system and it stalled.
These error messages are not needed. Remove them.
Link: https://lore.kernel.org/r/20220303140203.12642-1-sreekanth.reddy@broadcom.co... Suggested-by: Christoph Hellwig hch@lst.de Signed-off-by: Sreekanth Reddy sreekanth.reddy@broadcom.com Signed-off-by: Martin K. Petersen martin.petersen@oracle.com
We see this regression after upgrading from Linux 5.10 to 5.15 on our file servers with Broadcom/LSI SAS3008 PCI-Express Fusion-MPT SAS-3 (mpt3sas) – though luckily our systems do not stall/crash.
The commit message does not say anything about, what commit caused these error to be appearing – the log statements have been there since v4.20-rc1, if I am not mistaken, so it must be something else –, and also do not mention, why these log messages are not needed, but the new error condition is actually expected.
In the Canonical/Ubuntu bug tracker I found the explanation below [2].
- mpt3sas: Remove scsi_dma_map errors messages:
When driver set the DMA mask to 32bit then we observe that the SWIOTLB bounce buffers are getting exhausted quickly. For most of the IOs driver observe that scsi_dma_map() API returned with failure status and hence driver was printing below error message. Since this error message is getting printed per IO and if user issues heavy IOs then we observe that kernel overwhelmed with this error message. Also we will observe the kernel panic when the serial console is enabled. So to limit this issue, we removed this error message though this patch. "scsi_dma_map failed: request for 1310720 bytes!"
The Launchpad issue was created in March 2022, and the fixed Linux kernel package 5.15.0-53.59 for Ubuntu 22.04 released on November 15th, 2022.
Sreekanth, looking again, you are the patch author, one of the Broadcom maintainers (LSILOGIC MPT FUSION DRIVERS (FC/SAS/SPI)) and created the Launchpad bug report. I am surprised you didn’t get it backported upstream.
Kind regards,
Paul
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h... [2]: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1965927 "[Ubuntu 22.04] mpt3sas: Request to include latest bug fix patches"
On Mon, Jan 16, 2023 at 03:23:46PM +0100, Paul Menzel wrote:
Dear Linux folks,
Could you please apply commit 0c25422d34b4 (scsi: mpt3sas: Remove scsi_dma_map() error messages) to the 5.15.y series?
commit 0c25422d34b4726b2707d5f38560943155a91b80 Author: Sreekanth Reddy sreekanth.reddy@broadcom.com Date: Thu Mar 3 19:32:03 2022 +0530
scsi: mpt3sas: Remove scsi_dma_map() error messages When scsi_dma_map() fails by returning a sges_left value less than zero, the amount of logging produced can be extremely high. In a recent
end-user environment, 1200 messages per second were being sent to the log buffer. This eventually overwhelmed the system and it stalled.
These error messages are not needed. Remove them. Link:
https://lore.kernel.org/r/20220303140203.12642-1-sreekanth.reddy@broadcom.co... Suggested-by: Christoph Hellwig hch@lst.de Signed-off-by: Sreekanth Reddy sreekanth.reddy@broadcom.com Signed-off-by: Martin K. Petersen martin.petersen@oracle.com
We see this regression after upgrading from Linux 5.10 to 5.15 on our file servers with Broadcom/LSI SAS3008 PCI-Express Fusion-MPT SAS-3 (mpt3sas) – though luckily our systems do not stall/crash.
The commit message does not say anything about, what commit caused these error to be appearing – the log statements have been there since v4.20-rc1, if I am not mistaken, so it must be something else –, and also do not mention, why these log messages are not needed, but the new error condition is actually expected.
Now queued up, thanks.
greg k-h
[Cc: +loberman@redhat.com, +djeffery@redhat.com]
Am 16.01.23 um 15:23 schrieb Paul Menzel:
Dear Linux folks,
Could you please apply commit 0c25422d34b4 (scsi: mpt3sas: Remove scsi_dma_map() error messages) to the 5.15.y series?
commit 0c25422d34b4726b2707d5f38560943155a91b80 Author: Sreekanth Reddy sreekanth.reddy@broadcom.com Date: Thu Mar 3 19:32:03 2022 +0530
scsi: mpt3sas: Remove scsi_dma_map() error messages
When scsi_dma_map() fails by returning a sges_left value less than zero, the amount of logging produced can be extremely high. In a recent end-user environment, 1200 messages per second were being sent to the log buffer. This eventually overwhelmed the system and it stalled.
These error messages are not needed. Remove them.
Link: https://lore.kernel.org/r/20220303140203.12642-1-sreekanth.reddy@broadcom.co... Suggested-by: Christoph Hellwig hch@lst.de Signed-off-by: Sreekanth Reddy sreekanth.reddy@broadcom.com Signed-off-by: Martin K. Petersen martin.petersen@oracle.com
We see this regression after upgrading from Linux 5.10 to 5.15 on our file servers with Broadcom/LSI SAS3008 PCI-Express Fusion-MPT SAS-3 (mpt3sas) – though luckily our systems do not stall/crash.
The commit message does not say anything about, what commit caused these error to be appearing – the log statements have been there since v4.20-rc1, if I am not mistaken, so it must be something else –, and also do not mention, why these log messages are not needed, but the new error condition is actually expected.
In the Canonical/Ubuntu bug tracker I found the explanation below [2].
- mpt3sas: Remove scsi_dma_map errors messages:
When driver set the DMA mask to 32bit then we observe that the SWIOTLB bounce buffers are getting exhausted quickly. For most of the IOs driver observe that scsi_dma_map() API returned with failure status and hence driver was printing below error message. Since this error message is getting printed per IO and if user issues heavy IOs then we observe that kernel overwhelmed with this error message. Also we will observe the kernel panic when the serial console is enabled. So to limit this issue, we removed this error message though this patch. "scsi_dma_map failed: request for 1310720 bytes!"
The Launchpad issue was created in March 2022, and the fixed Linux kernel package 5.15.0-53.59 for Ubuntu 22.04 released on November 15th, 2022.
Sreekanth, looking again, you are the patch author, one of the Broadcom maintainers (LSILOGIC MPT FUSION DRIVERS (FC/SAS/SPI)) and created the Launchpad bug report. I am surprised you didn’t get it backported upstream.
Kind regards,
Paul
"[Ubuntu 22.04] mpt3sas: Request to include latest bug fix patches"
linux-stable-mirror@lists.linaro.org