linux-kernel - Re: [PATCH for-rc 5/9] RDMA/hns: Fix missing pagesize and alignment check in FRMR

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ccd00d9d-4cad-4b18-ae73-e618d669959e@linux.dev>
Date: Mon, 8 Jul 2024 15:57:49 +0800
From: Zhu Yanjun <yanjun.zhu@...ux.dev>
To: Junxian Huang <huangjunxian6@...ilicon.com>, jgg@...pe.ca, leon@...nel.org
Cc: linux-rdma@...r.kernel.org, linuxarm@...wei.com,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH for-rc 5/9] RDMA/hns: Fix missing pagesize and alignment
 check in FRMR

在 2024/7/8 10:44, Junxian Huang 写道:
> 
> 
> On 2024/7/7 17:16, Zhu Yanjun wrote:
>> 在 2024/7/5 16:59, Junxian Huang 写道:
>>> From: Chengchang Tang <tangchengchang@...wei.com>
>>>
>>> The offset requires 128B alignment and the page size ranges from
>>> 4K to 128M.
>>>
>>> Fixes: 68a997c5d28c ("RDMA/hns: Add FRMR support for hip08")
>>
>> https://patchwork.kernel.org/project/linux-rdma/patch/2eee7e35-504e-4f2a-a364-527e90669108@CMEXHTCAS1.ad.emulex.com/
>> In the above link, from Bart, it seems that FRMR is renamed to FRWR.
>> "
>> There are already a few drivers upstream in which the fast register
>> memory region work request is abbreviated as FRWR. Please consider
>> renaming FRMR into FRWR in order to avoid confusion and in order to
>> make it easier to find related code with grep in the kernel tree.
>> "
>>
>> So is it possible to rename FRMR to FRWR?
>>
> 
> I think the rename is irrelevant to this bugfix, and if it needs to be done,
> we'll need a single patch to rename all existing 'FRMR' in hns driver.
> 
> So let's leave it as is for now.

Exactly. FRMR is obsolete. We do need a single patch to rename all 
existing "FRMR" to "FRWR" in RDMA.

@Leon, please let me know your suggestions.
Thanks,

Zhu Yanjun

> 
> Thanks,
> Junxian
> 
>>> Signed-off-by: Chengchang Tang <tangchengchang@...wei.com>
>>> Signed-off-by: Junxian Huang <huangjunxian6@...ilicon.com>
>>> ---
>>>    drivers/infiniband/hw/hns/hns_roce_device.h | 4 ++++
>>>    drivers/infiniband/hw/hns/hns_roce_mr.c     | 5 +++++
>>>    2 files changed, 9 insertions(+)
>>>
>>> diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
>>> index 5a2445f357ab..15b3b978a601 100644
>>> --- a/drivers/infiniband/hw/hns/hns_roce_device.h
>>> +++ b/drivers/infiniband/hw/hns/hns_roce_device.h
>>> @@ -83,6 +83,7 @@
>>>    #define MR_TYPE_DMA                0x03
>>>      #define HNS_ROCE_FRMR_MAX_PA            512
>>> +#define HNS_ROCE_FRMR_ALIGN_SIZE        128
>>>      #define PKEY_ID                    0xffff
>>>    #define NODE_DESC_SIZE                64
>>> @@ -189,6 +190,9 @@ enum {
>>>    #define HNS_HW_PAGE_SHIFT            12
>>>    #define HNS_HW_PAGE_SIZE            (1 << HNS_HW_PAGE_SHIFT)
>>>    +#define HNS_HW_MAX_PAGE_SHIFT            27
>>> +#define HNS_HW_MAX_PAGE_SIZE            (1 << HNS_HW_MAX_PAGE_SHIFT)
>>> +
>>>    struct hns_roce_uar {
>>>        u64        pfn;
>>>        unsigned long    index;
>>> diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c
>>> index 1a61dceb3319..846da8c78b8b 100644
>>> --- a/drivers/infiniband/hw/hns/hns_roce_mr.c
>>> +++ b/drivers/infiniband/hw/hns/hns_roce_mr.c
>>> @@ -443,6 +443,11 @@ int hns_roce_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
>>>        struct hns_roce_mtr *mtr = &mr->pbl_mtr;
>>>        int ret, sg_num = 0;
>>>    +    if (!IS_ALIGNED(*sg_offset, HNS_ROCE_FRMR_ALIGN_SIZE) ||
>>> +        ibmr->page_size < HNS_HW_PAGE_SIZE ||
>>> +        ibmr->page_size > HNS_HW_MAX_PAGE_SIZE)
>>> +        return sg_num;
>>> +
>>>        mr->npages = 0;
>>>        mr->page_list = kvcalloc(mr->pbl_mtr.hem_cfg.buf_pg_count,
>>>                     sizeof(dma_addr_t), GFP_KERNEL);
>>