Hi Leon,

On 29.06.2021 10:40, Leon Romanovsky wrote:
> From: Maor Gottlieb <ma...@nvidia.com>
>
> orig_nents should represent the number of entries with pages,
> but __sg_alloc_table_from_pages sets orig_nents as the number of
> total entries in the table. This is wrong when the API is used for
> dynamic allocation where not all the table entries are mapped with
> pages. It wasn't observed until now, since RDMA umem who uses this
> API in the dynamic form doesn't use orig_nents implicit or explicit
> by the scatterlist APIs.
>
> Fix it by:
> 1. Set orig_nents as number of entries with pages also in
>     __sg_alloc_table_from_pages.
> 2. Add a new field total_nents to reflect the total number of entries
>     in the table. This is required for the release flow (sg_free_table).
>     This filed should be used internally only by scatterlist.
>
> Fixes: 07da1223ec93 ("lib/scatterlist: Add support in dynamic allocation of 
> SG table from pages")
> Signed-off-by: Maor Gottlieb <ma...@nvidia.com>
> Signed-off-by: Leon Romanovsky <leo...@nvidia.com>

This patch landed in linux-next 20210630 as commit a52724456928 
("lib/scatterlist: Fix wrong update of orig_nents"). It causes serious 
regression in DMA-IOMMU integration, which can be observed for example 
on ARM Juno board during boot:

Unable to handle kernel paging request at virtual address 00376f42a6e40454
Mem abort info:
   ESR = 0x96000004
   EC = 0x25: DABT (current EL), IL = 32 bits
   SET = 0, FnV = 0
   EA = 0, S1PTW = 0
   FSC = 0x04: level 0 translation fault
Data abort info:
   ISV = 0, ISS = 0x00000004
   CM = 0, WnR = 0
[00376f42a6e40454] address between user and kernel address ranges
Internal error: Oops: 96000004 [#1] PREEMPT SMP
Modules linked in:
CPU: 2 PID: 1 Comm: swapper/0 Not tainted 5.13.0-next-20210630+ #3585
Hardware name: ARM Juno development board (r1) (DT)
pstate: 80000005 (Nzcv daif -PAN -UAO -TCO BTYPE=--)
pc : __sg_free_table+0x60/0xa0
lr : __sg_free_table+0x7c/0xa0
..
Call trace:
  __sg_free_table+0x60/0xa0
  sg_free_table+0x1c/0x28
  iommu_dma_alloc+0xc8/0x388
  dma_alloc_attrs+0xcc/0xf0
  dmam_alloc_attrs+0x68/0xb8
  sil24_port_start+0x60/0xe0
  ata_host_start.part.32+0xbc/0x208
  ata_host_activate+0x64/0x150
  sil24_init_one+0x1e8/0x268
  local_pci_probe+0x3c/0xa0
  pci_device_probe+0x128/0x1c8
  really_probe+0x138/0x2d0
  __driver_probe_device+0x78/0xd8
  driver_probe_device+0x40/0x110
  __driver_attach+0xcc/0x118
  bus_for_each_dev+0x68/0xc8
  driver_attach+0x20/0x28
  bus_add_driver+0x168/0x1f8
  driver_register+0x60/0x110
  __pci_register_driver+0x5c/0x68
  sil24_pci_driver_init+0x20/0x28
  do_one_initcall+0x84/0x450
  kernel_init_freeable+0x31c/0x38c
  kernel_init+0x20/0x120
  ret_from_fork+0x10/0x18
Code: d37be885 6b01007f 52800004 540000a2 (f8656813)
---[ end trace 4ba4f0c9c48711a1 ]---
Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b

It looks that some changes to the scatterlist structures are missing 
outside of the lib/scatterlist.c.

For now I would suggest to revert this change.

> ---
>   include/linux/scatterlist.h |  8 ++++++--
>   lib/scatterlist.c           | 32 ++++++++------------------------
>   2 files changed, 14 insertions(+), 26 deletions(-)
>
> diff --git a/include/linux/scatterlist.h b/include/linux/scatterlist.h
> index 6f70572b2938..1c889141eb91 100644
> --- a/include/linux/scatterlist.h
> +++ b/include/linux/scatterlist.h
> @@ -35,8 +35,12 @@ struct scatterlist {
>   
>   struct sg_table {
>       struct scatterlist *sgl;        /* the list */
> -     unsigned int nents;             /* number of mapped entries */
> -     unsigned int orig_nents;        /* original size of list */
> +     unsigned int nents;             /* number of DMA mapped entries */
> +     unsigned int orig_nents;        /* number of CPU mapped entries */
> +     /* The fields below should be used internally only by
> +      * scatterlist implementation.
> +      */
> +     unsigned int total_nents;       /* number of total entries in the table 
> */
>   };
>   
>   /*
> diff --git a/lib/scatterlist.c b/lib/scatterlist.c
> index a59778946404..6db70a1e7dd0 100644
> --- a/lib/scatterlist.c
> +++ b/lib/scatterlist.c
> @@ -192,33 +192,26 @@ static void sg_kfree(struct scatterlist *sg, unsigned 
> int nents)
>   void __sg_free_table(struct sg_table *table, unsigned int max_ents,
>                    unsigned int nents_first_chunk, sg_free_fn *free_fn)
>   {
> -     struct scatterlist *sgl, *next;
> +     struct scatterlist *sgl, *next = NULL;
>       unsigned curr_max_ents = nents_first_chunk ?: max_ents;
>   
>       if (unlikely(!table->sgl))
>               return;
>   
>       sgl = table->sgl;
> -     while (table->orig_nents) {
> -             unsigned int alloc_size = table->orig_nents;
> -             unsigned int sg_size;
> +     while (table->total_nents) {
> +             unsigned int alloc_size = table->total_nents;
>   
>               /*
>                * If we have more than max_ents segments left,
>                * then assign 'next' to the sg table after the current one.
> -              * sg_size is then one less than alloc size, since the last
> -              * element is the chain pointer.
>                */
>               if (alloc_size > curr_max_ents) {
>                       next = sg_chain_ptr(&sgl[curr_max_ents - 1]);
>                       alloc_size = curr_max_ents;
> -                     sg_size = alloc_size - 1;
> -             } else {
> -                     sg_size = alloc_size;
> -                     next = NULL;
>               }
>   
> -             table->orig_nents -= sg_size;
> +             table->total_nents -= alloc_size;
>               if (nents_first_chunk)
>                       nents_first_chunk = 0;
>               else
> @@ -301,20 +294,11 @@ int __sg_alloc_table(struct sg_table *table, unsigned 
> int nents,
>               } else {
>                       sg = alloc_fn(alloc_size, gfp_mask);
>               }
> -             if (unlikely(!sg)) {
> -                     /*
> -                      * Adjust entry count to reflect that the last
> -                      * entry of the previous table won't be used for
> -                      * linkage.  Without this, sg_kfree() may get
> -                      * confused.
> -                      */
> -                     if (prv)
> -                             table->nents = ++table->orig_nents;
> -
> +             if (unlikely(!sg))
>                       return -ENOMEM;
> -             }
>   
>               sg_init_table(sg, alloc_size);
> +             table->total_nents += alloc_size;
>               table->nents = table->orig_nents += sg_size;
>   
>               /*
> @@ -385,12 +369,11 @@ static struct scatterlist *get_next_sg(struct sg_table 
> *table,
>       if (!new_sg)
>               return ERR_PTR(-ENOMEM);
>       sg_init_table(new_sg, alloc_size);
> +     table->total_nents += alloc_size;
>       if (cur) {
>               __sg_chain(next_sg, new_sg);
> -             table->orig_nents += alloc_size - 1;
>       } else {
>               table->sgl = new_sg;
> -             table->orig_nents = alloc_size;
>               table->nents = 0;
>       }
>       return new_sg;
> @@ -515,6 +498,7 @@ struct scatterlist *__sg_alloc_table_from_pages(struct 
> sg_table *sgt,
>               cur_page = j;
>       }
>       sgt->nents += added_nents;
> +     sgt->orig_nents = sgt->nents;
>   out:
>       if (!left_pages)
>               sg_mark_end(s);

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland

_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Reply via email to