kernel.git/crypto/scompress.c, branch linux-6.15.y

crypto: scompress - increment scomp_scratch_users when already allocated

2025-04-25T02:33:30Z

Commit ddd0a42671c0 only increments scomp_scratch_users when it was 0, causing a panic when using ipcomp: Oops: general protection fault, probably for non-canonical address 0xdffffc0000000000: 0000 [#1] SMP KASAN NOPTI KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007] CPU: 1 UID: 0 PID: 619 Comm: ping Tainted: G N 6.15.0-rc3-net-00032-ga79be02bba5c #41 PREEMPT(full) Tainted: [N]=TEST Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Arch Linux 1.16.3-1-1 04/01/2014 RIP: 0010:inflate_fast+0x5a2/0x1b90 [...] Call Trace: zlib_inflate+0x2d60/0x6620 deflate_sdecompress+0x166/0x350 scomp_acomp_comp_decomp+0x45f/0xa10 scomp_acomp_decompress+0x21/0x120 acomp_do_req_chain+0x3e5/0x4e0 ipcomp_input+0x212/0x550 xfrm_input+0x2de2/0x72f0 [...] Kernel panic - not syncing: Fatal exception in interrupt Kernel Offset: disabled ---[ end Kernel panic - not syncing: Fatal exception in interrupt ]--- Instead, let's keep the old increment, and decrement back to 0 if the scratch allocation fails. Fixes: ddd0a42671c0 ("crypto: scompress - Fix scratch allocation failure handling") Signed-off-by: Sabrina Dubroca Signed-off-by: Herbert Xu

crypto: scomp - Fix off-by-one bug when calculating last page

2025-04-23T01:32:57Z

Fix off-by-one bug in the last page calculation for src and dst. Reported-by: Nhat Pham Fixes: 2d3553ecb4e3 ("crypto: scomp - Remove support for some non-trivial SG lists") Signed-off-by: Herbert Xu

crypto: scomp - Fix wild memory accesses in scomp_free_streams

2025-04-12T01:33:09Z

In order to use scomp_free_streams to free the partially allocted streams in the allocation error path, move the alg->stream assignment to the beginning. Also check for error pointers in scomp_free_streams before freeing the ctx. Finally set alg->stream to NULL to not break subsequent attempts to allocate the streams. Fixes: 3d72ad46a23a ("crypto: acomp - Move stream management into scomp layer") Reported-by: syzkaller Co-developed-by: Kuniyuki Iwashima Signed-off-by: Kuniyuki Iwashima Co-developed-by: Herbert Xu Signed-off-by: Herbert Xu

crypto: scomp - Fix null-pointer deref when freeing streams

2025-04-09T13:28:31Z

As the scomp streams are freed when an algorithm is unregistered, it is possible that the algorithm has never been used at all (e.g., an algorithm that does not have a self-test). So test whether the streams exist before freeing them. Reported-by: Sourabh Jain Fixes: 3d72ad46a23a ("crypto: acomp - Move stream management into scomp layer") Signed-off-by: Herbert Xu Tested-by: Sourabh Jain Signed-off-by: Herbert Xu

crypto: scompress - Fix incorrect stream freeing

2025-03-21T09:39:06Z

Fix stream freeing crash by passing the correct pointer. Fixes: 3d72ad46a23a ("crypto: acomp - Move stream management into scomp layer") Signed-off-by: Herbert Xu

crypto: scompress - Fix scratch allocation failure handling

2025-03-21T09:39:05Z

If the scratch allocation fails, all subsequent allocations will silently succeed without actually allocating anything. Fix this by only incrementing users when the allocation succeeds. Fixes: 6a8487a1f29f ("crypto: scompress - defer allocation of scratch buffer to first use") Signed-off-by: Herbert Xu Acked-by: Ard Biesheuvel Signed-off-by: Herbert Xu

crypto: acomp - Add support for folios

2025-03-21T09:35:26Z

For many users, it's easier to supply a folio rather than an SG list since they already have them. Add support for folios to the acomp interface. Signed-off-by: Herbert Xu

crypto: scomp - Add chaining and virtual address support

2025-03-21T09:33:39Z

Add chaining and virtual address support to all scomp algorithms. Signed-off-by: Herbert Xu

crypto: scomp - Remove support for some non-trivial SG lists

2025-03-21T09:33:39Z

As the only user of acomp/scomp uses a trivial single-page SG list, remove support for everything else in preprataion for the addition of virtual address support. However, keep support for non-trivial source SG lists as that user is currently jumping through hoops in order to linearise the source data. Limit the source SG linearisation buffer to a single page as that user never goes over that. The only other potential user is also unlikely to exceed that (IPComp) and it can easily do its own linearisation if necessary. Also keep the destination SG linearisation for IPComp. Signed-off-by: Herbert Xu

crypto: scomp - Disable BH when taking per-cpu spin lock

2025-03-15T08:21:23Z

Disable BH when taking per-cpu spin locks. This isn't an issue right now because the only user zswap calls scomp from process context. However, if scomp is called from softirq context the spin lock may dead-lock. Signed-off-by: Herbert Xu